Actualités financières en direct

The AI bubble and it’s possible disintegration gives us some pretty grim insights into the question of what happens when the richest people in the world rebuild the economy in a way that it …

The AI bubble and it’s possible disintegration gives us some pretty grim insights into the question of what happens when the richest people in the world rebuild the economy in a way that it no longer needs working people, but still expects them to pay their bills. Reuters reported this week that China’s AI and semiconductor stock boom has suffered a sudden and staggering reversal: Their STAR50 index of Beijing’s AI and chip champions (like our NASDAQ) has plunged 40% in less than four months, wiping out roughly $480 billion in market value. Their robotics company Unitree's IPO just three months ago was oversubscribed 8,000 times by Chinese small investors and has lost half its value. Dong Baozhen, a Beijing fund manager, told Reuters: And now we see the bubble bursting. The same day (Monday) here in America Microsoft offered voluntary buyouts to 60,000 employees, about 27% of its global workforce, as it’s planning to replace them with AI. And when Anthropic (Claude) filed to go public, they disclosed an $80.7 billion net loss and some $657 billion in infrastructure deals with companies including Microsoft, Amazon, and SpaceX; fully $498 billion of that is noncancelable even if the company never uses the computing capacity it’s reserved. Things got even weirder on Wednesday, when Microsoft and Alphabet reported some striking “earnings” numbers. Alphabet burned through $9.1 billion in free cash flow for the second straight quarter. NPR’s 1A tallied up Monday’s AI deals alone at roughly $235 billion, nearly 1% of American GDP, all in just one single day. While Wall Street has been pouring massive sums into AI companies, data centers, chips, and power infrastructure, Economist Dean Baker reports at his weekly AI Bubble Monitor that CoreWeave’s debt (the company rents computing power to the AI giants) is now considered so deep in junk bond territory that they’re having to pay 16% interest on their long-term debt to get investors to bite. In 1965, Intel co-founder Gordon Moore postulated that every two years the number of transistors in an integrated circuit would double. Not only was his “Moore’s Law” right, but as the power of digital hardware increases, costs also reliably drop by a similar factor. Most of today’s frenzy of data center and power plant construction is based on the current state of the AI chip-and-software art, but Moore’s Law dictates that over time—and not a lot of time, probably just a matter of months or years at the most—the size and power needs of these AI data centers will decrease exponentially. Already, entire buildings filled with computers are on the verge of being replaced by a single “wafer” disc that will carry the computing power of billions of today’s red-hot chips. The Wall Street Journal broke the news with an article titled and subtitled: The Microchip Era Is About to End. The future is in wafers. Data centers will be the size of a box, not vast energy-hogging structures. Author George Gilder noted we’re not just on the verge of the breakthrough; it’s nearly in production: Cerebras of Palo Alto, Calif., used the concept in its WSE-3 wafer-scale engine. The WSE-3 boasts some four trillion transistors—14 times as many as Nvidia’s Blackwell chip—with 7,000 times the memory bandwidth. Cerebras inscribed the memory directly on to the wafer rather than relegating it to distant chips and chiplets in high-bandwidth memory mazes. The company stacked up its wafer-scale engines 16-fold, thereby reducing a data center to a small box with 64 trillion transistors. A data center—a 700,000 square-foot building drawing gigawatts of electricity and chugging millions of gallons of water for cooling—replaced by “a small box” that could be powered by rooftop solar and a good battery bank. Meanwhile, multiple companies have rolled out AI products that can run on a desktop PC, eliminating the need for data centers altogether. Perplexity introduced one just last month with the headline: Perplexity Computer now runs fully local on Mac. Ya think that may have an impact on our economy? Places booming today as AI and other data facilities and the power plants and transmission lines to feed them are being built—if Moore’s Law applies to data centers as it has proven to apply to all things digital from computers to cell phones to TVs and satellites—will soon look like Flint, Michigan when Ronald Reagan’s free trade policies began to seriously bite in the 1990s. It sure looks like we’re seeing a repeat of previous bubbles and bursts ranging from Dutch tulips to the 19th century railroads to this century’s dot-com and home mortgage collapses. And the people running the show seem just fine with creating a world where the rich get mindbogglingly richer while working people and 401k investors are out on the street. JD Vance’s sugar daddy, Peter Thiel, wrote in a 2009 essay for the Cato Institute that “I no longer believe that freedom and democracy are compatible.” Marc Andreesen, the billionaire who helped Elon recruit young men for DOGE to take apart our government, complained on a podcast this year that government workers have both civil service protections and public-sector unions, giving them “two layers of insulation against ever getting fired.” Unions, of course, are actual democracy in the workplace that generations of workers have literally fought and died for, but these guys don’t seem to appreciate or even understand that. Meanwhile, OpenAI CEO Sam Altman wrote in a 2021 essay that as AI advances, “even more power will shift from labor to capital.” His warning points out how easily AI could dramatically increase the power of people who own businesses and tech while gutting the bargaining power of people who work for a living. And, indeed, that’s already happening. Like Microsoft this week, Amazon has cut roughly 30,000 corporate jobs since last October, all after CEO Andy Jassy warned his employees that AI would “reduce our total corporate workforce.” One AI-related layoff tracker counts nearly 119,000 layoffs so far this year tied explicitly to AI automation, roughly four and a half times last year’s total. And now the robotics business is exploding, with AI built into local machines that can assemble cars, sew clothing, clean streets, and drive you around town. It’s opened an entirely new industry: outfitting humans with sensors to track their movements to train robots to replace them. So, if machines replace our workers, who’s going to have enough money to buy the goods and services the machines are putting out? I grew up in a union family; my dad worked in a tool-and-die shop in Lansing, Michigan and was able to raise four boys with a three-bedroom home, a new car every few years, and had a decent retirement. Henry Ford increased his workers’ pay so they’d be able to buy the cars they were making. This isn’t rocket science. Workers are also customers, taxpayers, homeowners, and parents; they’re the ultimate economic base of their communities. If AI destroys their purchasing power, we’re gnawing away at the foundation of America’s historic economic greatness. We’ve seen versions of this before. Reagan deregulated the Savings & Loan industry leading to predatory behavior that produced a massive crash; it cost taxpayers over $100 billion to clean up the mess (a trillion in today’s money). But Reagan also sent those banksters to prison: His regulators made more than 30,000 criminal referrals to the Justice Department, which won more than 1,000 felony convictions. Since then, times have changed; it appears that the morbidly rich have acquired near-total immunity, no matter how badly they screw up things, because five corrupt Republicans on the Supreme Court legalized billionaires and corporations owning judges and politicians. Congress and Bill Clinton deregulated our banks in the late 1990s and the bill came due in 2008 when it all fell apart and millions of Americans lost their homes, families watched retirement accounts evaporate, and workers who’d done absolutely nothing wrong lost their livelihoods. Nepo-baby George W. Bush chose to hold nobody accountable, and even let the banksters keep their billion-dollar bonuses pulled out just before all the dominoes collapsed. He made zero criminal referrals against the big banks, in a crisis that was 70 times larger than Reagan’s S&L mess. It appears that the AI gods tech bros think that’s the new normal and are betting on getting the same or better from the Trump regime. After all, in his first term he put “Foreclosure King” Steve Mnuchin, who’d thrown an estimated 36,000 Californians out of their homes during the 2008 crash, in charge of the entire economy as Treasury Secretary. And this time the stakes are even larger than the banking system’s freeze-up; AI is now deeply entangled with America’s military, intelligence agencies, national security infrastructure, and the rivalry with China. And in their arrogance, they’re saying the quiet part out loud: Last November, OpenAI’s chief financial officer, Sarah Friar, told a Wall Street Journal conference that her company wanted a federal “backstop” for the financing behind its data centers. When a reporter asked whether she meant a government guarantee, she answered, “Exactly.” And pouring gasoline on this particular bonfire, President Donald Trump’s One Big Beautiful Billionaire’s Bill lets these companies write off the full cost of their data center hardware the year they buy it rather than the normal depreciation schedules every other industry in America has to stick to. Perhaps that’s because the Trump family is so well-invested now in this, and he’s depending on tech bro billions to keep Republicans in office; Elon alone is spending almost a half-billion so far this year to that end. Meanwhile, ordinary Americans are already paying. Residential electricity prices rose 6.9% last year, more than double the inflation rate, and Goldman Sachs’ analysts say they expect data centers to drive a massive 40% of electricity demand growth through the end of the decade. The 13 mid-Atlantic and Midwest states have been hit with $23 billion in consumer electricity price increases through at least 2028, driven by AI data centers. A homeowner in Manassas, Virginia opened a $281 electric bill in January, up from about $100 the month before. If today’s boom goes bust, who’s going to be left stuck with the bills for the half-built transmission lines, data centers, and power plants? Apparently it won’t be the morbidly rich tech bro masters of the universe, any more than it was Bush’s billionaire bankers. None of this is to diminish the power and importance of AI. But the internet was real, too, and look what happened in the 1999-2002 period when that industry took down people’s retirement accounts while creating billionaires out of Zuck, Bezos, and Gates. In 1983, Reagan instructed his Department of Justice, Federal Trade Commission, and Securities and Exchange Commision to stop enforcing the Sherman Anti-Trust and other anti-monopoly laws, leading to a decade of “Mergers & Acquisitions Mania” as was documented in Michael Douglas’ “greed is good” movie, Wall Street. Add that to Reagan decriminalizing share buybacks and you have the perfect storm of perverse incentives that end up with doors falling off airplanes, workers and communities being screwed, and a new low standard of crappy customer service—all fed by a lack of competition. And we see that consolidation now happening in AI. This is what happens when the people who own the machines accumulate more economic—and, thus, political—power than the people who depend on wages. It’s the old question: “Is the economy here to serve us, or are we here to serve the small number of mostly white men who control the economy?” The AI revolution is going to force that debate back into the open. The ultimate betrayal will be if AI succeeds in making America vastly wealthier while leaving working Americans with less economic security, less political power, and with the bill for another generation of financial speculation wiping out our retirement accounts. If that happens, we need to be ready to force the issue of the concentration of wealth and power that grew up around it. We failed with the internet giants in 2000 and the banksters in 2008. We need to prepare—politically and economically—for the next cycle of boom and bust that’s increasingly seeming inevitable.

Binance said Sunday it had resumed withdrawals for some major cryptocurrencies such as Bitcoin and Ethereal after more than $6 billion in assets were stolen from the cryptocurrency exchange. Binance is the world's largest …

Binance said Sunday it had resumed withdrawals for some major cryptocurrencies such as Bitcoin and Ethereal after more than $6 billion in assets were stolen from the cryptocurrency exchange. Binance is the world's largest crypto exchange by daily trading volu… Binance said Sunday it had resumed withdrawals for some major cryptocurrencies such as Bitcoin and Ethereal after more than $6 billion in assets were stolen from the cryptocurrency exchange.Binance i…

In this photo illustration, the Binance logo is displayed on a screen on June 6, 2023 in San Anselmo, California. Binance said Sunday that it had resumed withdrawals after a major hack that affected …

In this photo illustration, the Binance logo is displayed on a screen on June 6, 2023 in San Anselmo, California. Binance said Sunday that it had resumed withdrawals after a major hack that affected more than $6 billion in funds. Binance, which is the world’s largest crypto exchange by daily trading volume, had suspended user withdrawals after the attack. “First batch network now opened in phase one of withdrawals is $BTC bitcoin network,” Binance said on X. The company said in an earlier update that it was preparing to resume the first phase of withdrawals between 0600 and 0900 GMT on Sunday. “We are taking a phased approach by prioritising tokens based on the past week’s data of user demand, so that each network and asset pair can be restored carefully, with stability and security checks completed before access is reopened,” it said. Binance said Saturday that it has set aside more than $6.34bn in company funds to cover losses in the hack, which is thought to be the biggest-ever crypto heist. “User funds remain safe. Binance will absorb the financial impact of the stolen funds, not our users,” it said on X. News of the cyberattack caused the prices of most major cryptocurrencies to fall sharply on Friday. AFP

Binance to Resume Customer Withdrawals in Phases After HackBloomberg.com Binance Resumes Withdrawals After Suffering Biggest-Ever Crypto HackWSJ Binance Re-Enables Withdrawals on BTC, ETH, BSC, XRP and Base NetworksBinance Binance reopens XRP, ETH, and USDC …

Binance to Resume Customer Withdrawals in Phases After HackBloomberg.com Binance Resumes Withdrawals After Suffering Biggest-Ever Crypto HackWSJ Binance Re-Enables Withdrawals on BTC, ETH, BSC, XRP and Base NetworksBinance Binance reopens XRP, ETH, and USDC w… The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

Today, Sunday, November 1, 2026, the highest CD rate among our verified partners is 4.50%, offered by Capital One on its 11-month CD and by Synchrony Bank on its 15-month CD. The Federal Reserve …

Today, Sunday, November 1, 2026, the highest CD rate among our verified partners is 4.50%, offered by Capital One on its 11-month CD and by Synchrony Bank on its 15-month CD. The Federal Reserve held interest rates steady on October 28 after raising the feder… Today, Sunday, November 1, 2026, the highest CD rate among our verified partners is 4.50%, offered by Capital One on its 11-month CD and by Synchrony Bank on its 15-month CD.The Federal Reserve held …

The AI trade just had its scariest October in years. The rest of the stock market barely noticed. The biggest AI and chip stocks have shed about $5 trillion in market value since Oct. …

The AI trade just had its scariest October in years. The rest of the stock market barely noticed. The biggest AI and chip stocks have shed about $5 trillion in market value since Oct. 6, by Yahoo Finance calculations. Nvidia (NVDA) and Alphabet (GOOGL) have each given up roughly $835 billion in market cap, and Intel (INTC) lost nearly half its value in October. Everything else? The other 625 large US stocks in that tally added about $1.3 trillion. And while the S&P 500 (^GSPC) slipped this week, 9 of its 11 sectors rose about 1% or more, something that hadn't happened in any week since at least 1999, by Yahoo Finance calculations. Yahoo Finance Jeff Hirsch, editor of the Stock Trader's Almanac, has a name for this: "Octoberphobia." And he's buying the sell-off. "I did buy some today... this is an opportunity here," Hirsch told Yahoo Finance on Friday. "The damage isn't as bad as the headlines are." Hirsch also has the calendar on his side. October selloffs like this one often lead into what the Almanac calls the sweet spot of the presidential election cycle, which runs from the fourth quarter of a midterm year through the following June. History is on the buyers' side From the last October close of a midterm year to the end of the following June, the S&P 500 has risen 19 times out of 19 since 1950, with a median gain of about 15%. Chip stocks have been 8-for-8 since 1994, with a median gain of about 22%. Yahoo Finance The ride wasn't always smooth. Chips fell another 11% or so before turning higher in 2002 and 2018, and eight cycles is a relatively small sample. Chips just sprang a possible bear trap The iShares Semiconductor ETF (SOXX) tumbled about 17% in October, yet it's still up more than 50% this year. On Friday, it dipped as low as $460, below the $464 low from July 29 that marked this summer's bottom. Then it closed at $472, back above that line. That's the look of a bear trap — a breakdown that reverses, catching traders betting on more losses. Chip stocks briefly broke below their July low Friday, then closed back above it. · Yahoo Finance AlphaSpace Hirsch likes the timing. "Historically, seasonally, it sets up great for this period of time, the best months of the year for semiconductors," he said. His firm placed a new buy in the sector on Friday. Bulls first need to reclaim about $524, the top of last week's bounce, then the early-October high near $596. Losing the July low would put the February-April range near $370 in play. That's about 21% below Friday's close. The case for waiting Not everyone is loading up the shopping cart. BofA's Michael Hartnett said it's "risk-off until global rate hike ends," pointing to 42 central bank rate hikes worldwide over the past six months. The Fed held on Wednesday, but two officials voted to hike. Hartnett's line in the sand for chips is $520 on the VanEck Semiconductor ETF (SMH), which closed Friday near $522. On SOXX, that works out to about $469, just above the July low it barely held.

For decades, voters have trusted the GOP to keep the gross domestic product growing based on one misleading idea: They know that the GOP is the party of business, so they just assume surely …

For decades, voters have trusted the GOP to keep the gross domestic product growing based on one misleading idea: They know that the GOP is the party of business, so they just assume surely the pro-business party is better for the economy. But voters are turning away from that, and the reason why could mark a seismic change in American politics: Now, they aren’t so sure that what’s good for businesses is good for them. They’re looking for candidates with not just answers but different kinds of answers than they’ve heard before. If Democrats win next Tuesday and actually deliver next year, it could usher in a new era of policy that finally closes the door on the neoliberalism that has dominated for nearly half a century. The idea that what’s good for the private sector is good for working people has deep roots in the GOP, but in our time, it comes from Ronald Reagan, whose economic policies pointed to government spending as the cause of inflation, claiming it crowded out innovation and cost workers jobs and opportunities. He slashed the federal government’s spending and said he would focus welfare programs on only “the truly needy,” the definition of which was always shrinking. He promised that public revenue would grow even as the tax rates that individuals paid fell, because of overall growth in the economy. “Our aim is to increase our national wealth so all will have more, not just redistribute what we already have, which is just a sharing of scarcity,” Reagan said soon after being elected in 1981. “We can begin to reward hard work and risk-taking, by forcing this government to live within its means.”

Binance said Sunday it had resumed withdrawals for some major cryptocurrencies such as Bitcoin and Ethereal after more than $6 billion in assets were stolen from the cryptocurrency exchange. Binance is the world's largest …

Binance said Sunday it had resumed withdrawals for some major cryptocurrencies such as Bitcoin and Ethereal after more than $6 billion in assets were stolen from the cryptocurrency exchange. Binance is the world's largest crypto exchange by daily trading volu… Binance said Sunday it had resumed withdrawals for some major cryptocurrencies such as Bitcoin and Ethereal after more than $6 billion in assets were stolen from the cryptocurrency exchange.Binance i…

CabalSpy ID — SocialFi API & SDK for Python Official Python SDK (sync and async) for CabalSpy ID. CabalSpy ID is the SocialFi API for crypto apps: one SocialFi integration gives your platform login …

CabalSpy ID — SocialFi API & SDK for Python Official Python SDK (sync and async) for CabalSpy ID. CabalSpy ID is the SocialFi API for crypto apps: one SocialFi integration gives your platform login with X, wallets verified by a tiny on-chain payment (no wallet connect, no signatures), reviewed KOL badges, live trades and PnL of your users and a drop-in SocialFi widget for token theses — on Solana, BNB Chain, Base, Ethereum and Robinhood Chain. Your platform adds "Login with CabalSpy ID" and receives the user's X profile, verified wallets, KOL badge and full trading stats — PnL, win rate, trades, holdings — for copy trading, leaderboards and social features. Website · Docs · Sign in · API dashboard pip install cabalspy-id The live stream needs one extra dependency: pip install "cabalspy-id[stream]" CabalSpyIDAuth / AsyncCabalSpyIDAuth OpenID Connect for your server: PKCE, code exchange, refresh, revoke, userinfo, ID token verification (ES256 / JWKS) CabalSpyID / AsyncCabalSpyID Data API client for https://api-id.cabalspy.xyz/v1 , typed responses, pagination, live trade stream, white-label platform users, trading competitions, callouts (on request), the Trade Feed and copy trading ThesisClient / AsyncThesisClient public token thesis API with your public client_id and a user's access token, plus live() updates verify_webhook Standard Webhooks signature verification Sync and async, fully typed ( py.typed , TypedDict responses). Python 3.10+. Everything for developers — instance (client_id), client secrets, redirect URIs, button builder, webhooks, API keys — lives in the API dashboard, tab CabalSpy ID. For React and the browser, see the TypeScript SDK. What can you build Social trading feed — every trade of your users live, with their X profile and PnL — every trade of your users live, with their X profile and PnL Leaderboards — rank your users by PnL, volume, win rate or trades — rank your users by PnL, volume, win rate or trades Copy trading — follow verified wallets of real, identified traders; every trade of a copied trader arrives as a ready-to-execute instruction sized by the user's settings — follow verified wallets of real, identified traders; every trade of a copied trader arrives as a ready-to-execute instruction sized by the user's settings Token theses — holders-only opinions on a token with the author's real position, as a SocialFi widget or your own UI — holders-only opinions on a token with the author's real position, as a SocialFi widget or your own UI Trader profiles — X profile, verified wallets, holdings, PnL calendar — X profile, verified wallets, holdings, PnL calendar KOL badges — reviewed badges for influential traders, shown wherever they sign in — reviewed badges for influential traders, shown wherever they sign in White-label — keep your own login, import your users and their wallets, run trading competitions — keep your own login, import your users and their wallets, run trading competitions Callouts (on request) — your users call tokens; every call is stamped and tracked, with a score, a callers leaderboard and callout competitions Packages One codebase, one version line. The focused packages re-export parts of cabalspy-id , so they always behave exactly the same. Package npm PyPI What it contains cabalspy-id cabalspy-id cabalspy-id everything below socialfi-api (repo) socialfi-api socialfi-api the SocialFi API: Data API, Login with CabalSpy ID, trade stream, webhooks (= cabalspy-id/api , cabalspy_id.api ) socialfi-thesis (repo) socialfi-thesis socialfi-thesis token theses: Data API thesis methods, public thesis client, theses stream (= cabalspy-id/thesis , cabalspy_id.thesis ) socialfi-widget (repo) socialfi-widget – the browser SocialFi widgets: login button and theses widget, vanilla and React (= cabalspy-id/widget ) Modules Import Contents cabalspy_id everything cabalspy_id.api CabalSpyID , AsyncCabalSpyID , CabalSpyIDStream , CalloutsStream , CopyTradingStream , TradeFeedStream , CabalSpyIDAuth , verify_webhook , errors and Data API types cabalspy_id.thesis CabalSpyID (thesis methods), ThesisClient , AsyncThesisClient , ThesisLive , ThesesStream , thesis types The objects are the same as in cabalspy_id . Login button Put the official button on your page and point it at your login route: < div data-csid-login data-client-id = "csid_..." data-href = "/login" ></ div > < script src = "https://id.cabalspy.xyz/widget.js" async ></ script > Signed in, mark the person so the button shows their avatar menu: data-user="<sub>" data-logout-href="/logout" . Design overrides: data-theme , data-color , data-text-color , data-shape , data-size , data-label , data-full-width , data-show-logo . Server-side login (OpenID Connect) Authorization code flow with PKCE S256 (required for every client), state and nonce. from cabalspy_id import CabalSpyIDAuth , is_limited auth = CabalSpyIDAuth ( "csid_..." , # your instance's client_id client_secret = os . environ [ "CSID_CLIENT_SECRET" ], # csid_sk_... (omit for public clients) redirect_uri = "https://your.app/callback" , # registered in the dashboard ) # GET /login req = auth . create_authorization_url ( scope = [ "openid" , "profile" , "wallets" , "offline_access" ]) session [ "csid" ] = req . to_dict () # state, nonce, code_verifier, redirect_uri return redirect ( req . url ) # GET /callback result = auth . handle_callback ( request . url , ** session . pop ( "csid" )) # checks error, state and iss, exchanges the code, verifies the ID token and its nonce if is_limited ( result . claims ): ... # over your plan's user limit: anonymous sub "anon_...", no profile or wallet data else : user = auth . userinfo ( result . tokens [ "access_token" ]) # x profile, wallets, onchain Method discover() the discovery document (cached) create_authorization_url(scope=, prompt=, redirect_uri=, state=, nonce=) returns AuthorizationRequest(url, state, nonce, code_verifier, redirect_uri) handle_callback(url, state=, nonce=, code_verifier=) full callback handling, returns LoginResult(tokens, claims) exchange_code(code=, code_verifier=, redirect_uri=) POST /oauth/token , authorization_code refresh(refresh_token) POST /oauth/token , refresh_token. Tokens rotate: store the new one; reusing an old one revokes the chain revoke(token) POST /oauth/revoke userinfo(access_token) GET /oauth/userinfo verify_id_token(id_token, nonce=, max_age=) ES256 signature via JWKS, iss, aud, azp, exp, iat, nonce verify_access_token(access_token) for access tokens your frontend sends to your API ( typ: at+jwt ) AsyncCabalSpyIDAuth has the same methods as coroutines. Scopes: openid profile wallets onchain thesis offline_access ( thesis : post theses and replies, react and report in the user's name). Client authentication: client_secret_basic (default with a secret), client_secret_post , none . Access and ID tokens live 15 minutes; refresh tokens 30 days of inactivity, 90 days absolute. Browser apps (public clients) let the widget run PKCE itself; it hands your page an idToken . Send it to your backend and verify it there: auth = CabalSpyIDAuth ( "csid_..." ) claims = auth . verify_id_token ( id_token_from_the_browser ) Users over the plan's limit can always sign in. Their token carries an anonymous sub ( anon_... ), limited: true and reason: "plan_user_limit" , and no profile or wallet data, until the platform upgrades. is_limited(claims) checks both. A runnable server is in examples/login_server.py. Users without X People can sign in to CabalSpy ID with X, Telegram or Google. Accounts without X can do everything except earn the KOL badge, and connecting X later fills in their profile. For them: Field With X Without X preferred_username (claims) / username X handle, e.g. alice tg.<telegram name> (Telegram) or id.<hex> (Google) — always contains a dot x (userinfo, /v1/profile ) X profile / stats None twitter (profiles, trading profile blocks, leaderboards) https://x.com/<handle> None telegram (profiles, trading profile blocks, leaderboards) https://t.me/<name> or "" https://t.me/<name> or "" name , picture / image_url from X from Telegram or Google profile_url ( https://id.cabalspy.xyz/u/<handle> ) works for everyone. Thesis authors use the same handles. from cabalspy_id import has_x , handle_kind has_x ( claims ) # False for Telegram/Google sign-ins handle_kind ( "tg.alice" ) # "telegram" x_link = profile [ "twitter" ] # None without X: never build x.com links from tg./id. handles Data API from cabalspy_id import CabalSpyID csid = CabalSpyID () # reads CABALSPY_ID_API_KEY, then CABALSPY_API_KEY me = csid . me () # instance, plan, rate limit, users with data # The person who just signed in: look them up by the `sub` of their ID token profile = csid . profile ( csid_id = claims [ "sub" ]) wallet = next ( w for w in profile [ "wallets" ] if "solana" in w [ "chains" ]) # Full trading profile for a period, same numbers as the CabalSpy API t = csid . tracker ( blockchain = "solana" , address = wallet [ "address" ], period = "7d" ) print ( t [ "period_stats" ][ "realized_pnl_usd" ], t [ "period_win_rate_distribution" ][ "win_rate_percentage" ]) # Your users ranked by PnL board = csid . leaderboard ( blockchain = "solana" , period = "7d" , sort = "pnl_usd" , limit = 25 ) Async: from cabalspy_id import AsyncCabalSpyID async with AsyncCabalSpyID () as csid : users = await csid . users ( kol = True ) The API key is the same key as the CabalSpy API, with the per-key toggle "CabalSpy ID" enabled on the API keys page. A key sees its instance only: users that signed in to its apps, within the plan's user limit. scope="global" (every CabalSpy ID user) is Enterprise only. Method Endpoint health() GET /v1/health (no key needed) me() GET /v1/me users(limit=, kol=, cursor=, scope=) GET /v1/users profile(username= | csid_id= | address=) GET /v1/profile wallet_lookup(address) GET /v1/wallets/lookup tracker(blockchain=, address=, period=) GET /v1/wallets/tracker history(blockchain=, address=, limit=, cursor=) GET /v1/wallets/history holdings(blockchain=, address=) GET /v1/wallets/holdings pnl_calendar(blockchain=, address=) GET /v1/wallets/pnl_calendar leaderboard(blockchain=, period=, sort=, limit=, cursor=, scope=) GET /v1/wallets/leaderboard latest_transactions(blockchain=, limit=) GET /v1/transactions/latest Chains: solana , bnb , base , eth , rh . Periods: 6h , 1d , 7d , 30d . Leaderboard sorts: volume , pnl , pnl_usd , win_rate , trades . Responses are plain dicts typed with TypedDict s ( Profile , TrackerResponse , LeaderboardResponse , …). If your account has several CabalSpy ID instances, pass instance="csid_..." (the client_id). Pagination Iterators follow next_cursor for you (async generators on AsyncCabalSpyID ): for user in csid . iter_users ( kol = True ): print ( user [ "username" ]) for trade in csid . iter_history ( blockchain = "solana" , address = address ): print ( trade [ "tx_signature" ]) for entry in csid . iter_leaderboard ( blockchain = "base" , period = "30d" ): print ( entry [ "rank" ]) For whole pages with metadata: csid.pages("/users", {"limit": 100}) yields Envelope s, and csid.get_raw(path, params) returns one. Errors Every failure is a subclass of CabalSpyIDError carrying .code , .status , .request_id , .docs (a link to the error's documentation), and for validation errors .parameter and .allowed . from cabalspy_id import NotFoundError , PlanUserLimitError , RateLimitError try : csid . profile ( csid_id = claims [ "sub" ]) except PlanUserLimitError : show_upgrade_notice () # anon_ sub: no data until you upgrade except NotFoundError : pass # not one of your users except RateLimitError as err : print ( err . retry_after ) Class Status Codes BadRequestError 400, 422 missing_parameter , invalid_parameter , instance_required , wallet_required , invalid_address , invalid_return_url , cannot_add_self , no_price (422) AuthenticationError 401 missing_api_key PermissionError 403 invalid_api_key , product_not_enabled , no_instance , scope_not_allowed , claim_disabled , feature_not_enabled , kol_only , not_a_holder , invalid_account , posting_paused , consent_required , sandbox_user , trading_private PlanUserLimitError 403 plan_user_limit NotFoundError 404 profile_not_found , wallet_not_found , instance_not_found , platform_user_not_found , verification_not_found , competition_not_found , callout_not_found , trader_not_found , not_in_trade_feed , not_found ConflictError 409 account_claimed , last_wallet , too_many_wallets , wallet_already_verified , kol_badge , x_required , kol_not_eligible , competition_limit , competition_closed , competition_full , late_join_closed , already_called , copy_managed_elsewhere , already_reported , trade_feed_full RateLimitError 429 rate_limit_exceeded , rate_limited (also the Trade Feed's 120 changes per hour; not retried), too_many_open_verifications , callout_cooldown ServerError 5xx internal_error , service_unavailable APIConnectionError – network failure, timeout OAuthError 4xx OAuth errors from the login endpoints ( invalid_grant , invalid_client , …) TokenVerificationError – ID / access token failed verification WebhookVerificationError – webhook signature or timestamp invalid PlatformImportError – bulk_import_platform_users stopped; carries result and remaining 429 , 5xx and network errors are retried with exponential backoff (default 2 retries); Retry-After wins. Writes (POST) are retried only after the API-wide rate limit ( rate_limit_exceeded ), since they did not run. callout_cooldown is never retried: the wait ( retry_after ) can be hours. The rate limit of the last call is in csid.last_rate_limit ( X-RateLimit-Limit/Remaining/Reset ). Keep your own login. Your platform imports its users — by your own user id ( external_id ) — together with the wallets they already have (Privy, Dynamic, any provider), and then uses their trades, PnL, leaderboards and trading competitions through the API. No CabalSpy ID login, no wallet connect: imported wallets count right away (unless you set wallet_trust="verified_only" ), and the on-chain payment verification stays available as a badge. from cabalspy_id import CabalSpyID , PlanUserLimitError csid = CabalSpyID () # One user: create or update (a new user needs 1-10 wallets) user = csid . upsert_platform_user ( "user_42" , wallets = [ "So1..." , { "address" : "0xabc..." , "chains" : [ "base" ], "source" : "privy" }], display_name = "Alice" , ) # Everyone else: any number of users, sent in chunks of 1,000 result = csid . bulk_import_platform_users ( { "external_id" : u . id , "wallets" : u . wallet_addresses , "display_name" : u . name } for u in all_users ) print ( result [ "summary" ]) # {"created": ..., "updated": ..., "unchanged": ..., "rejected": ...} # A profile page in one call, and your own leaderboard profile = csid . platform_user_profile ( "user_42" , period = "7d" ) board = csid . platform_leaderboard ( blockchain = "solana" , period = "7d" , sort = "pnl_usd" ) for entry in board [ "leaderboard" ]: print ( entry [ "rank" ], entry [ "external_id" ], entry [ "period_stats" ][ "realized_pnl_usd" ]) # A trading competition between your users comp = csid . create_competition ( name = "Weekly PnL" , chain = "solana" , metric = "pnl_usd" , duration_hours = 168 , enrollment = "all_users" ) standings = csid . competition_standings ( comp [ "id" ]) Every platform-users method that takes a user accepts the external_id positionally or csid_id= (exactly one). The rest of the API also accepts external_id in place of csid_id (pass it with get_raw / post_raw ), and every platform user carries its csid_id for the other SDK methods. A full tour is in examples/white_label.py . Bulk import. import_platform_users(users) is one call (at most 1,000 users). bulk_import_platform_users(users) takes any number: it sends chunks of at most 1,000 users and under the server's 1 MB body limit, retries a chunk refused with 429 ( Retry-After wins), and sends users rejected with rate_limited (too many new users per minute) again on their own once the minute is over. A chunk that succeeded is never sent again. Results come back aggregated, one per user in input order. If anything else stops it, PlatformImportError carries result (what was imported) and remaining (pass it again to resume). Each user is handled on its own on the server: a bad item or the plan's user limit ( plan_user_limit ) never undoes the others. Method Endpoint upsert_platform_user(external_id, wallets=, display_name=, avatar_url=) POST /v1/platform-users import_platform_users(users) / bulk_import_platform_users(users) POST /v1/platform-users/import platform_user(external_id | csid_id=) GET /v1/platform-users update_platform_user(external_id, display_name=, avatar_url=) POST /v1/platform-users/update platform_users(limit=, cursor=) / iter_platform_users() GET /v1/platform-users/list delete_platform_user(external_id) POST /v1/platform-users/delete add_platform_user_wallets(external_id, wallets=) POST /v1/platform-users/wallets/add delete_platform_user_wallet(external_id, address=) POST /v1/platform-users/wallets/delete start_wallet_verification(external_id, chain=, address=) POST /v1/platform-users/wallets/verify wallet_verification(id) / wallet_verification_qr(id) (SVG) / cancel_wallet_verification(id) /v1/platform-users/wallets/verification[/qr|/cancel] platform_user_profile(external_id, period=) GET /v1/platform-users/profile platform_leaderboard(blockchain=, period=, sort=, limit=, cursor=, trust=) / iter_platform_leaderboard() GET /v1/platform-users/leaderboard platform_settings() / update_platform_settings(...) GET / POST /v1/platform-users/settings platform_inbound() / enable_platform_inbound() / disable_platform_inbound() /v1/platform-users/inbound[/enable|/disable] platform_user_session(external_id) POST /v1/platform-users/session (verify widget, 15 min) platform_user_claim_link(external_id, return_url=) POST /v1/platform-users/claim-link connect_platform_user_x(external_id, return_url=) / disconnect_platform_user_x(external_id) POST /v1/platform-users/x/connect , /x/disconnect link_platform_user_telegram / platform_user_telegram / unlink_platform_user_telegram /v1/platform-users/telegram[/link|/unlink] platform_user_kol / request_platform_user_volume_check / apply_platform_user_kol /v1/platform-users/kol[/volume-check|/apply] Settings ( update_platform_settings ): white_label , data_scope ( csid : part of the CabalSpy ID network, default; platform : only you see your users), wallet_trust and leaderboard_trust ( all or verified_only ), claim_links , telegram_alerts . Wallets carry a status : verified , active (unverified, counts), needs_verification , verified_elsewhere or pending . Inbound URL. enable_platform_inbound() returns a secret URL (shown once) that your platform or its auth provider calls on sign-up; users are then created like upsert_platform_user . Formats: csid , privy , dynamic , generic . Trading competitions Individual competitions between your platform users, at most 5 scheduled + running at a time per account. A competition starts at starts_at (or at once) once min_participants joined, then runs duration_hours . Metrics: pnl_usd , pnl , volume_usd , volume , trades , win_rate . Traders below min_trades / min_volume_usd are listed but not ranked. Rules can change at any time, also while running; every change is recorded ( changes , rules_version ). Method Endpoint create_competition(name=, chain=, metric=, duration_hours=, ...) POST /v1/competitions competitions(status=, limit=, offset=) / iter_competitions() GET /v1/competitions competition(id) GET /v1/competitions/detail update_competition(id, note=, **fields) POST /v1/competitions/update cancel_competition(id) POST /v1/competitions/cancel join_competition(id, external_id) / leave_competition(id, external_id) POST /v1/competitions/join , /leave competition_participants(id, limit=, offset=) / iter_competition_participants(id) GET /v1/competitions/participants competition_standings(id, limit=, offset=) / iter_competition_standings(id) GET /v1/competitions/standings The competition routes page by offset ; the iter_* helpers walk them until pagination.total . Webhooks: platform_user.created , platform_user.claimed , platform_user.x_connected , wallet.added , verification.expired , and per instance competition.started , competition.updated , competition.standings , competition.ended ( PLATFORM_WEBHOOK_EVENTS , COMPETITION_WEBHOOK_EVENTS ). Callouts (on request) Callouts are an extra for CabalSpy ID instances: they are switched on per instance on request (contact CabalSpy). Until then every callout call raises PermissionError with the code feature_not_enabled ; csid.me()["features"] lists "callouts" once they are on. A user calls a token: its price and market cap are stamped at that moment and tracked. First 24 hours: read every minute (current and peak multiple, milestones 2x, 3x, 5x, 10x, ...), then the day is scored on price candles ( first_day , webhook callout.scored ). read every minute (current and peak multiple, milestones 2x, 3x, 5x, 10x, ...), then the day is scored on price candles ( , webhook ). After that: read every 5 minutes while the token is active, plus each day's candles. Tracking stops after 24 hours in a row below a $10k market cap ( tracking.below_10k_since , then callout.final ). A callout's score is score_x , its peak multiple since the call. On top: stats per user and period, a callers leaderboard, a live stream, webhooks and callout competitions. from cabalspy_id import CabalSpyID , ConflictError , RateLimitError csid = CabalSpyID () try : c = csid . create_callout ( "user_42" , chain = "solana" , mint = MINT , note = "Undervalued, team ships" ) print ( c [ "at_call" ][ "market_cap_usd" ], c [ "score_x" ]) except ConflictError : # already_called: one call per token and 24 hours pass except RateLimitError as err : # callout_cooldown: the instance's cooldown, never retried print ( "next callout in" , err . retry_after , "seconds" ) for c in csid . iter_callouts ( chain = "solana" , mint = MINT ): # chart markers, newest first print ( c [ "called_at" ], c [ "author" ][ "name" ], c [ "peak" ][ "x" ], c [ "tracking" ][ "phase" ]) stats = csid . callout_stats ( "user_42" , hit_x = 2 ) # per period: 6h, 1d, 7d, 30d print ( stats [ "periods" ][ "7d" ][ "avg_peak_x" ], stats [ "periods" ][ "7d" ][ "hit_rate" ]) board = csid . callout_leaderboard ( period = "7d" , sort = "avg_peak_x" , min_callouts = 3 ) for e in board [ "leaderboard" ]: print ( e [ "rank" ], e [ "author" ][ "name" ], e [ "avg_peak_x" ], e [ "best_call" ][ "symbol" ]) csid . update_callout_settings ( who = "all" , require_holding = True , cooldown_minutes = 360 ) Method Endpoint create_callout(external_id | csid_id=, chain=, mint=, note=) POST /v1/callouts remove_callout(id, reason=) POST /v1/callouts/remove (moderation) callouts(chain=, mint=, external_id=, csid_id=, since=, until=, limit=, cursor=) / iter_callouts() GET /v1/callouts callout(id) GET /v1/callouts/item callout_stats(external_id | csid_id=, hit_x=) GET /v1/callouts/stats callout_leaderboard(period=, chain=, sort=, min_callouts=, hit_x=, limit=, cursor=) / iter_callout_leaderboard() GET /v1/callouts/leaderboard callout_settings() / update_callout_settings(who=, require_holding=, cooldown_minutes=) GET / POST /v1/callouts/settings callouts_stream(chain=, mint=) WS /v1/callouts/stream A callout ( Callout ): at_call (price and market cap at the call), now and peak (price, market cap, multiple x , time), score_x , status ( tracking or final ), tracking ( active , phase = first_day / extended / stopped , next_check_at , below_10k_since , daily_checks , stopped_at , stop_reason ) and, once the first day is scored, first_day ( kind = candles or live , peak_x , close_x , minutes_to_peak , flags ). Stats and leaderboard. Per period (the callouts made in the last 6h / 1d / 7d / 30d): callouts , avg_peak_x , median_peak_x , best_peak_x , sum_peak_x , hits and hit_rate (percent of callouts whose peak reached hit_x , default 2), plus avg_now_x and tracking in the stats. The leaderboard sorts by avg_peak_x (default), best_peak_x , sum_peak_x , hit_rate , callouts or median_peak_x ; each entry has best_call . Settings. who ( all or kol_only ), require_holding (only holders may call a token, default on), cooldown_minutes (0-10080 between two callouts of one user, default 360). Errors on create_callout : kol_only , not_a_holder , posting_paused ( PermissionError ), already_called ( ConflictError ), callout_cooldown ( RateLimitError with retry_after ), no_price ( BadRequestError , 422). Live stream (async, needs cabalspy-id[stream] ): async with AsyncCabalSpyID () as csid : async with csid . callouts_stream ( chain = "solana" ) as cs : # mint= one token (needs chain) async for ev in cs : if ev [ "event" ] == "callout.milestone" : print ( ev [ "data" ][ "token" ][ "symbol" ], ev [ "data" ][ "milestone_x" ], "x" ) Events: callout.created , callout.update (a new price: price_usd , market_cap_usd , x , peak_x ), callout.milestone (the callout plus milestone_x ), callout.scored , callout.final , callout.removed , plus ready , heartbeat , pong , open , close , reconnecting , error . feature_not_enabled and invalid_api_key are fatal; too_many_connections (at most 5 per key) reconnects with backoff. Callout competitions. create_competition(kind="callouts", ...) ranks your users on the callouts they make during the competition: metrics avg_peak_x , best_peak_x , sum_peak_x , hit_rate , callouts ; min_callouts to be ranked, hit_x the multiple that counts as a hit, and chain="all" to count every chain. Standings carry the callout stats and best_call . The kind cannot change later; min_trades / min_volume_usd belong to trading competitions. comp = csid . create_competition ( name = "Caller of the week" , kind = "callouts" , chain = "all" , metric = "avg_peak_x" , duration_hours = 168 , min_callouts = 3 , hit_x = 2 , enrollment = "all_users" ) A full tour is in examples/callouts.py . Trade Feed A user's Trade Feed is the traders they follow: every trade and thesis of those traders, alerts filtered per trader, and the copy settings. Read and manage it from your backend in the user's name: a person who signed in to your app with the scope trade_feed (else PermissionError consent_required ), or one of your platform users, within your plan's user limit. Every method takes the user as external_id (positional, your own user id) or csid_id= ; a trader is a handle (with or without @) or a csid_id, or trader_external_id= for one of your platform users. from cabalspy_id import CabalSpyID , ConflictError , PermissionError csid = CabalSpyID () # Add @MazeCCC to user_42's Trade Feed: alerts for buys from 0.5 SOL, no sells try : r = csid . add_trade_feed_trader ( "user_42" , trader = "@MazeCCC" , min_sol = 0.5 , notify_sells = False ) print ( r [ "created" ], r [ "settings" ][ "min_native" ]) # True on HTTP 201, False when already there except PermissionError as err : # consent_required, trading_private print ( err . code ) except ConflictError : # trade_feed_full: 300 traders pass for item in csid . iter_trade_feed ( "user_42" , filter = "buys" ): # newest first t = item [ "trade" ] print ( item [ "trader" ][ "username" ], t [ "side" ], t [ "symbol" ], t [ "amount_usd" ], t [ "label" ]) for entry in csid . iter_trade_feed_traders ( "user_42" ): print ( entry [ "trader" ][ "username" ], entry [ "settings" ][ "muted" ], entry [ "settings" ][ "copy" ][ "enabled" ]) csid . remove_trade_feed_trader ( "user_42" , trader = "MazeCCC" ) # also ends copying Method Endpoint trade_feed(external_id | csid_id=, filter=, limit=, cursor=) / iter_trade_feed() GET /v1/trade-feed trade_feed_traders(external_id | csid_id=, limit=, cursor=) / iter_trade_feed_traders() GET /v1/trade-feed/traders add_trade_feed_trader(external_id | csid_id=, trader= | trader_external_id=, notify_buys=, notify_sells=, notify_theses=, min_sol=, min_bnb=, min_eth=, muted=) POST /v1/trade-feed/traders remove_trade_feed_trader(external_id | csid_id=, trader= | trader_external_id=) POST /v1/trade-feed/traders/delete trade_feed_notifications(external_id | csid_id=, limit=, cursor=) / iter_trade_feed_notifications() GET /v1/trade-feed/notifications trade_feed_stream() WS /v1/trade-feed/stream The feed ( TradeFeedItem , newest first; filter all , buys , sells or theses ; limit 1-100, default 30): trade items ( type trade : trader , trade with side , label ( new , add , partial , close , sell ), sold_percent , amount_native in currency , amount_usd , price_usd , pnl_usd / pnl_percent for sells with a known entry, wallet , tx_signature and explorer links) and thesis items ( type thesis , id h + the thesis id). An item's id is the same in the feed, the webhook, the stream and the notifications. Traders who keep their trading private are left out. Traders ( TradeFeedTrader , most recently added first; limit 1-300, default 100; the envelope's pagination has total and max ): trader , trading_public , added_at , in_trade_feeds and settings ( notify_buys , notify_sells , notify_theses , min_native {"SOL", "BNB", "ETH"} , muted , and copy , the CopySettings ). add_trade_feed_trader uses the user's default alert settings unless you pass others; for a trader already there it updates what you pass (None = unchanged). min_* are 0 to 1,000,000 (0 = every trade). Notifications ( TradeFeedNotification , limit 1-100, default 50): every alert that passed the user's filters, with id , event ( trade_feed.trade / trade_feed.thesis ), a ready-made text , created_at and the item as data . Useful as an inbox when you do not use webhooks. Errors. BadRequestError cannot_add_self ; PermissionError consent_required , trading_private ; PlanUserLimitError plan_user_limit (an anon_ id); NotFoundError trader_not_found , platform_user_not_found ; ConflictError trade_feed_full (300 traders); RateLimitError rate_limited (more than 120 changes of a user's Trade Feed in an hour; not retried). Live stream (async, needs cabalspy-id[stream] ) and webhooks trade_feed.trade / trade_feed.thesis ( TRADE_FEED_WEBHOOK_EVENTS ) carry every new trade and thesis that concerns your users, once per event (not per user): data is the item plus recipients (csid_ids of your users whose filters let it through) and recipients_external (your own ids of those who are platform users). async with AsyncCabalSpyID () as csid : async with csid . trade_feed_stream () as tf : async for ev in tf : # {"event": "trade_feed.trade", "id": ..., "data": ...} for user in ev [ "data" ][ "recipients_external" ]: await notify ( user , ev [ "data" ]) Events: trade_feed.trade , trade_feed.thesis , plus ready , heartbeat , pong , open , close , reconnecting , error . invalid_api_key is fatal; too_many_connections (at most 5 per key) reconnects with backoff. Events made while disconnected are not replayed: read the notifications or the feed to catch up. A full tour is in examples/trade_feed.py . Copy trading Every trade of a trader one of your users copies becomes a ready-to-execute instruction for that user, sized by their copy settings for that trader. CabalSpy ID never holds keys and never executes: your app executes the instruction with the user's wallet and reports what happened. It needs the user's consent with the scope trade_feed (else PermissionError consent_required ); a copied trader is in the user's Trade Feed. from cabalspy_id import CabalSpyID , ConflictError csid = CabalSpyID () # Copy @ctkol for user_42: $25 per buy, at most $20, buys and sells, 1.5 % price limit try : s = csid . update_copy_trading_settings ( "user_42" , trader = "ctkol" , enabled = True , mode = "fixed_usd" , amount_usd = 25 , max_usd = 20 , sides = "both" , slippage_bps = 150 ) except ConflictError : # copy_managed_elsewhere: another platform copies this trader for the user pass for s in csid . copy_trading_settings ( "user_42" ): # every trader in the Trade Feed print ( s [ "trader" ][ "username" ], s [ "copy" ][ "enabled" ], s [ "managed_by_other_platform" ]) for i in csid . iter_copy_instructions ( "user_42" , status = "pending" ): # newest first a = i [ "action" ] sig = execute ( i [ "csid_id" ], a ) # your swap, before i["expires_at"] csid . report_copy_instruction ( i [ "id" ], status = "executed" , tx_signature = sig , executed_amount_usd = a [ "amount_usd" ]) Method Endpoint copy_trading_settings(external_id | csid_id=, trader=, trader_external_id=) GET /v1/copy-trading/settings (a list, or one with a trader) update_copy_trading_settings(external_id | csid_id=, trader= | trader_external_id=, enabled=, mode=, amount_usd=, percent=, max_usd=, sides=, slippage_bps=) POST /v1/copy-trading/settings copy_instructions(external_id | csid_id=, status=, limit=, cursor=) / iter_copy_instructions() GET /v1/copy-trading/instructions copy_instruction(id) GET /v1/copy-trading/instructions/item report_copy_instruction(id, status=, tx_signature=, executed_amount_usd=, note=) POST /v1/copy-trading/instructions/status copy_trading_stream() WS /v1/copy-trading/stream Settings ( CopySettings , per user and trader): enabled , mode ( fixed_usd : every buy is amount_usd , default 50; percent : percent % of the trader's buy in USD, default 100, at most 1000), max_usd (cap per buy; max_usd=None removes it), sides ( buys or both , default both), slippage_bps (1-5000, default 300). Only the settings you pass change; the trader (handle or csid_id, or trader_external_id for one of your platform users) is added to the Trade Feed if needed (HTTP 201). Copying a trader is switched on by one platform at a time: only its apps get the instructions, so a user on two platforms is never copied twice ( managed_by_other_platform , copy_managed_elsewhere ). An instruction ( CopyInstruction ): id , status , created_at , expires_at (120 seconds later), csid_id , trader , action ( blockchain , mint , symbol , side ; buys: amount_usd , amount_native in currency and max_price_usd ; sells: sell_percent of the copied position and min_price_usd ; price_usd , slippage_bps ), source_trade (the trader's trade), settings (as they were), links , and once reported tx_signature , executed_amount_usd , note . Buys under $1 are not sent. A sell is only sent after a copied buy of that token that was not reported failed or skipped; a closing sell is sell_percent 100. A pending instruction past expires_at reads as expired (it may still be reported executed). Reporting. status is executed , failed or skipped and final: the same report twice is fine, another one raises ConflictError already_reported . Reports are not retried after a server or network error. Live stream (async, needs cabalspy-id[stream] ), webhook copy_trade.instruction ( COPY_TRADING_WEBHOOK_EVENTS ) and REST carry the same instructions; the stream and the webhook add external_id : async with AsyncCabalSpyID () as csid : async with csid . copy_trading_stream () as cs : async for ev in cs : # {"event": "copy_trade.instruction", "id": ..., "data": ...} i = ev [ "data" ] await execute_and_report ( csid , i ) Events: copy_trade.instruction , plus ready , heartbeat , pong , open , close , reconnecting , error . invalid_api_key is fatal; too_many_connections (at most 5 per key) reconnects with backoff. A full tour is in examples/copy_trading.py . Token theses A thesis is a short opinion on a token (up to 1000 characters), optionally tagged bull or bear. Only holders post theses: one of the author's verified wallets bought the token and still holds it. Every thesis shows the author's real position — what they bought when posting and their live PnL now (or "sold"); authors who keep their trading private show only that they hold. Anyone signed in can reply and react ( like plus 🔥 🚀 💎 👀 😂 🤝 💀 🐂 🐻). The same theses appear in the widget, through the Data API and on every platform. Writing needs the scope thesis in your login ( scope=["openid", "profile", "thesis"] ). Scope of reads. Theses are shared across platforms. Every read ( list_theses , thesis_summary , latest_theses , ThesisClient , the widget, the live sockets) takes scope : global (default, every platform) or instance (only theses written on your instance). In the Data API your key decides the instance; the public API and widget use your client_id . Your public client_id is required for the public thesis API ( ThesisClient , AsyncThesisClient , ThesisLive ) and the theses widget. It is your instance's client_id ( csid_... ), shown in the dashboard next to the API key. It is public, not a secret and not the API key (the API key must never reach a browser). The SDK sends it as the X-CSID-Client-Id header on every call to https://id.cabalspy.xyz/thesis/v1/... and as client_id on the live socket URL ( wss://id.cabalspy.xyz/thesis/v1/ws/{chain}/{mint}?client_id=... ). Without it ThesisClient() raises AuthenticationError with code == "client_id_required" . Browser calls must come from one of the instance's allowed origins or verified domains (instances in dev mode also allow http://localhost:* ). The Data API ( CabalSpyID , API key) is unchanged. Widget < div data-csid-thesis data-chain = "solana" data-mint = "..." data-client-id = "csid_..." data-theme = "dark" ></ div > < script src = "https://id.cabalspy.xyz/widget.js" async ></ script > < script > CSID . thesis . setAccessToken ( accessToken ) // optional: the signed-in person writes right away </ script > Data API (server to server) Reading works with any key, for every token and user. Writing happens in the name of one of your users: pass their csid_id (the sub of their ID token); they must have signed in to your app with the scope thesis and be within your plan's user limit. theses = csid . list_theses ( blockchain = "solana" , mint = mint , sort = "new" ) counts = csid . thesis_summary ( blockchain = "solana" , mint = mint ) # theses, bull, bear, authors, token ok = csid . thesis_eligibility ( blockchain = "solana" , mint = mint , csid_id = claims [ "sub" ]) if ok [ "can_post" ]: t = csid . post_thesis ( blockchain = "solana" , mint = mint , body = "Accumulating here." , stance = "bull" , csid_id = claims [ "sub" ]) Method Endpoint list_theses(blockchain=, mint=, sort=, stance=, limit=, cursor=, csid_id=) / iter_theses() GET /v1/theses thesis_summary(blockchain=, mint=) GET /v1/theses/summary thesis(id, csid_id=) GET /v1/theses/item thesis_replies(id, limit=, cursor=) / iter_thesis_replies() GET /v1/theses/replies latest_theses(blockchain=, users="all" | "instance", limit=, cursor=) GET /v1/theses/latest user_theses(csid_id= | username=) GET /v1/theses/user thesis_eligibility(blockchain=, mint=, csid_id=) GET /v1/theses/eligibility post_thesis(blockchain=, mint=, body=, stance=, csid_id=) POST /v1/theses reply_to_thesis(thesis_id=, body=, csid_id=) POST /v1/theses/replies react_to_thesis(thesis_id=, emoji=, on=, csid_id=) POST /v1/theses/reactions report_thesis(thesis_id=, reason=, note=, csid_id=) POST /v1/theses/report delete_thesis(thesis_id=, csid_id=) POST /v1/theses/delete | thesis_notifications(csid_id) | GET /v1/theses/notifications | | appeal_thesis_decision(csid_id, notification_id, message) | POST /v1/theses/appeal | Every thesis carries reacted_by (the latest reactions with who reacted) and, for theses, latest_replies (a preview of the newest replies). Moderation. Posts hidden after reports, pauses and moderator decisions reach the author as notifications ( id , kind , text , data , created_at , read , appealable , appeal ); appealable ones can be appealed within 30 days with 10-1000 characters. Through the Data API in your user's name: thesis_notifications(csid_id) / appeal_thesis_decision(...) ; with the user's token: ThesisClient.notifications() , mark_notifications_read() , appeal(id, message) . ThesisClient.me() also returns notices ( unread , latest , url ) for a badge. Thesis error codes: not_a_holder , consent_required , scope_missing , plan_user_limit ( PlanUserLimitError ), posting_paused , reporting_paused , sandbox_user (403, PermissionError ); invalid_body , duplicate (400); not_found (404); rate_limited (429). Writes are never retried after a server or network error, so a post can't be duplicated; only the API's own rate limit is retried. With data-client-id the widget loads the design saved in the dashboard; data-theme , data-scope , data-sort , data-font , data-radius , data-bg , data-card , data-text , data-muted , data-border , data-accent , data-accent-text , data-bull , data-bear , data-show-header , data-show-sentiment , data-compact and data-height override it. Theses stream (every plan) Live theses for the tokens you choose, e.g. to put markers on a chart. Up to 200 tokens per connection (or all), on every plan; subscriptions are sent again after every reconnect. Async, needs the stream extra. async with AsyncCabalSpyID () as csid : ts = csid . theses_stream ( scope = "global" ) # or "instance" @ts . on ( "thesis" ) def on_thesis ( t ): add_marker ( t [ "blockchain" ], t [ "mint" ], t ) await ts . connect () # returns on "ready" await ts . subscribe ([{ "blockchain" : "solana" , "mint" : MINT_A }, { "blockchain" : "base" , "mint" : TOKEN_B }]) await ts . unsubscribe ([{ "blockchain" : "base" , "mint" : TOKEN_B }]) # await ts.subscribe_all() / await ts.unsubscribe_all() await ts . run_forever () # or iterate {"event": ..., "data": ...} async with csid . theses_stream () as ts : await ts . subscribe_all () async for ev in ts : print ( ev [ "event" ], ev [ "data" ]) Events: ready , subscribed , thesis , thesis.reply , thesis.reaction , thesis.removed , heartbeat , reconnecting , close , error . A refused connection ( invalid_api_key , …) is fatal; a bad frame ( too_many_tokens , an invalid address) is reported as error and the connection stays. Also over the trade stream (Pro plan): csid.stream(events=["theses"], blockchain="solana", mint=mint) emits thesis , thesis.reply , thesis.reaction and thesis.removed (or iterate stream.theses() ). Your users' activity also arrives as webhooks: thesis.created ( event["data"]["thesis"] ), thesis.reply ( event["data"]["reply"] ) and thesis.reaction ( thesis_id , chain , mint , emoji , on , reactions ). Public thesis API: ThesisClient https://id.cabalspy.xyz/thesis/v1 needs no API key, only your public client_id (required): reading needs nothing else, writing uses the user's access token (scope thesis ). from cabalspy_id import ThesisClient # client_id: your instance's PUBLIC client_id from the dashboard (not the API key); it also credits # posts to your app. Add scope="instance" to show only theses written on your platform. with ThesisClient ( access_token = result . tokens [ "access_token" ], client_id = "csid_..." ) as theses : page = theses . list ( "solana" , mint , sort = "top" ) # items, pagination, token, summary if theses . me ( "solana" , mint )[ "can_post" ]: theses . post ( "solana" , mint , body = "Strong community, still early." , stance = "bull" ) theses . react ( page . items [ 0 ][ "id" ], "🚀" ) Methods: list , iterate , summary , get , replies , by_user , me , post , reply , react , report , delete , notifications , mark_notifications_read , appeal , live , set_access_token . AsyncThesisClient has the same methods as coroutines. Live updates for one token (async): live = ThesisClient ( client_id = "csid_..." ) . live ( "solana" , mint ) # ?client_id=... (also takes scope=) live . on ( "thesis" , lambda t : print ( "new" , t [ "body" ])) # also: reply, reaction, removed await live . connect () await live . run_forever () client_id errors use the SDK's exception classes, so check err.code (docs: https://docs.cabalspy.xyz/cabalspy-id/errors#<code> ): Code HTTP Live socket close Exception Meaning client_id_required 401 4401 AuthenticationError no client_id sent invalid_client 401 4401 AuthenticationError unknown or inactive client_id origin_not_allowed 403 4403 PermissionError the browser's Origin is not one of the instance's allowed origins / verified domains rate_limited 429 4429 RateLimitError ( retry_after == 60 ) too many requests per client or per IP ThesisLive does not reconnect after 4401 or 4403 ( run_forever() raises the error); after 4429 it waits 60 s before reconnecting. Live stream Every trade of your users the moment it is indexed, with the trader's profile (Pro plan and up). Async only, needs pip install "cabalspy-id[stream]" . async with AsyncCabalSpyID () as csid : stream = csid . stream ( blockchain = "solana" ) # omit blockchain for every chain # events=["trades"] (default), ["theses"] or both; mint= theses of one token @stream . on ( "trade" ) def on_trade ( t ): print ( t [ "trader" ][ "username" ], t [ "transaction_type" ], t [ "token_symbol" ], t [ "value_usd" ]) await stream . connect () # returns once the server confirmed the subscription await stream . run_forever () # until stream.close() or a fatal error # or iterate async with csid . stream () as stream : async for trade in stream : print ( trade ) The server sends a heartbeat every 30 seconds. The client reconnects with exponential backoff after a dropped connection or when heartbeats stop; configuration errors ( invalid_api_key , plan_required , scope_not_allowed ) are fatal and stop reconnecting. Events: trade , thesis , thesis.reply , thesis.reaction , thesis.removed , subscribed , heartbeat , open , close , reconnecting , error ; handlers may be sync or async. Webhooks Deliveries follow the Standard Webhooks spec ( webhook-id , webhook-timestamp , webhook-signature , secret whsec_... ) and are retried for about 2 days. Events: wallet.verified , wallet.removed , profile.updated , kol.granted , consent.revoked , thesis.created , thesis.reply , thesis.reaction — only for users within the plan's limit. White-label adds platform_user.created , platform_user.claimed , platform_user.x_connected , wallet.added , verification.expired and the instance-level competition.started , competition.updated , competition.standings , competition.ended (these carry competition_id instead of sub ). Callouts add callout.created , callout.milestone , callout.scored , callout.final ( data.callout ; CALLOUT_WEBHOOK_EVENTS ). Copy trading adds copy_trade.instruction (one per user and copied trade; data is the instruction plus external_id ; COPY_TRADING_WEBHOOK_EVENTS ). The Trade Feed adds trade_feed.trade and trade_feed.thesis (one per trade or thesis, not per user; data is the feed item plus recipients and recipients_external ; TradeFeedTradeEvent / TradeFeedThesisEvent ; TRADE_FEED_WEBHOOK_EVENTS ). from cabalspy_id import verify_webhook # pass the RAW body, exactly as received event = verify_webhook ( request . body , request . headers , os . environ [ "CSID_WEBHOOK_SECRET" ]) if event [ "type" ] == "consent.revoked" : delete_user_data ( event [ "data" ][ "sub" ]) Timestamps older than 5 minutes are rejected ( tolerance= ). During a secret roll several signatures are sent; any match is accepted. Dedupe on event["id"] , it is the same on every retry. sign_webhook() builds test requests for your receiver. Configuration CabalSpyID ( api_key = None , # env fallback: CABALSPY_ID_API_KEY, then CABALSPY_API_KEY instance = None , # client_id, when the account has several instances base_url = "https://api-id.cabalspy.xyz/v1" , stream_url = "wss://api-id.cabalspy.xyz/v1/stream" , callouts_stream_url = "wss://api-id.cabalspy.xyz/v1/callouts/stream" , copy_trading_stream_url = "wss://api-id.cabalspy.xyz/v1/copy-trading/stream" , trade_feed_stream_url = "wss://api-id.cabalspy.xyz/v1/trade-feed/stream" , timeout = 30.0 , max_retries = 2 , headers = {}, http_client = None , # bring your own httpx.Client / httpx.AsyncClient ) CabalSpyIDAuth ( "csid_..." , client_secret = None , # csid_sk_..., server apps only redirect_uri = None , issuer = "https://id.cabalspy.xyz" , token_endpoint_auth_method = None , # client_secret_basic | client_secret_post | none clock_tolerance = 60 , timeout = 15.0 , http_client = None , ) Keep the API key and the client secret on your server; read them from the environment or a secret manager. Pricing Per month and registrable domain. Everyone can always sign in; above the limit the platform receives no data for the extra users until it upgrades. Plan Price / month Users with data API rate Free $0 50 1 req/s Starter $49 200 5 req/s Pro $149 500 10 req/s + live stream Business $499 1,000 25 req/s Enterprise custom unlimited custom + SLA scope=global Yearly billing is 20% off. Card or crypto at apidashboard.cabalspy.xyz/billing/csid. Docs docs.cabalspy.xyz/cabalspy-id · Discovery: id.cabalspy.xyz/.well-known/openid-configuration Related License MIT

Polling is looking grim for Republicans ahead of the U.S. midterm elections on November 3, due in large part to deep pessimism among the public about the economy. At first glance, this might seem …

Polling is looking grim for Republicans ahead of the U.S. midterm elections on November 3, due in large part to deep pessimism among the public about the economy. At first glance, this might seem surprising. With the exception of inflation, many economic meas… Polling is looking grim for Republicans ahead of the U.S. midterm elections on November 3, due in large part to deep pessimism among the public about the economy.At first glance, this might seem surp…