Model or dataset
alsk1992/CloddsBot avatar
alsk1992/CloddsBot

CloddsBot (Clodds): a self-hosted AI trading terminal for prediction markets, crypto and futures

Open Source AI trading agent that operates autonomously across 1000+ markets - Polymarket, Kalshi, Binance, Hyperliquid, Solana DEXs, 5 EVM chains. Scans for edge, executes instantly, manages risk while you sleep. Agent commerce protocol for machine-to-machine payments. Self-hosted. Built on Claude.

2,734 stars331 forksTypeScriptMIT

At a glance

What is it?
CloddsBot by alsk1992 is an MIT-licensed, self-hosted AI trading terminal built on Claude. You chat with it through one of 21 messaging platforms or a built-in WebChat, and it trades across prediction markets, crypto spot, and perpetual futures. It is powerful and broad, and it puts real money at risk, so treat it accordingly.
Who is it for?
Consider CloddsBot only if you understand automated trading and are willing to risk real money on an autonomous agent. Install the current GitHub release, run clodds onboard, and start with WebChat and read-only exploration before adding any exchange keys.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 4 days ago.
What is it written in?
Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.

DEEP OPEN-SOURCE ANALYSIS

What CloddsBot is and the risk to state first

CloddsBot, branded Clodds (Claude plus Odds), is a personal AI trading terminal for prediction markets, crypto spot, perpetual futures with margin, token launches and Bittensor subnet mining. You run it on your own machine, chat with it through any of 21 messaging platforms or a built-in WebChat, and it trades across 10 prediction markets and 7 futures exchanges, with Solana and EVM-chain integrations, all through natural conversation. It is powered by Claude and ships with, by the README's own count, well over 100 trading strategies plus whale tracking, arbitrage detection, copy trading and DCA bots.

The risk has to come before the features. This is software that executes real financial trades autonomously, with up to 200x leverage on some futures venues. Automated trading can lose money quickly, high margin magnifies that, and prediction and crypto markets are volatile and, in some jurisdictions, legally restricted. The README is a feature list, not investment advice, and nothing here is either; anyone running Clodds is responsible for the trades it makes and for whether using it is legal where they live.

Installing and onboarding

The requirement is firm: Node.js 22 or newer, since Node 20 is unsupported and dependency installation may fail. The README installs the current release straight from GitHub, because the old npm package is no longer maintained:

bash
npm install -g https://github.com/alsk1992/CloddsBot/releases/latest/download/clodds.tgz --loglevel=error
clodds onboard

The onboarding wizard walks through the API key, a messaging channel and starting the gateway, and WebChat then opens at http://localhost:18789/webchat. There is a from-source alternative:

bash
git clone https://github.com/alsk1992/CloddsBot.git && cd CloddsBot
npm install && cp .env.example .env
# Add ANTHROPIC_API_KEY to .env
npm run build && npm start

Only ANTHROPIC_API_KEY is strictly required to run; a messaging channel is recommended to interact. Crucially, trading credentials are optional and separate, so you can run and explore Clodds without wiring any exchange keys at all, which is the sensible way to start.

The CLI and the built-in WebChat

Clodds is driven by a small CLI. Beyond clodds onboard, there is clodds start to launch the gateway, clodds repl for an interactive shell, clodds doctor for diagnostics, clodds secure to harden security, clodds locale set to change language, and clodds mcp to run an MCP server (with clodds mcp install to auto-configure Claude Desktop or Claude Code). That MCP path is notable: it means Clodds can be exposed as a tool to Claude Desktop or Claude Code rather than only used standalone.

The built-in WebChat is more capable than a bare console. It has a Claude-style sidebar with Chats, Projects, Artifacts and Code tabs, auto-extracts artifacts and code blocks, stores every message in an append-only database table with no cap so long histories load quickly, and does context compacting by summarising older messages so the AI keeps a compressed recap plus the last 20 messages. It is SQLite-backed and persists across restarts, with a nine-language selector. For a self-hosted tool, having a full local chat UI with no third-party dependency is a real convenience.

The breadth of markets and channels

The scope is unusually wide. On messaging, 21 platforms are supported, including Telegram, Discord, WhatsApp, Slack, Teams, Signal, Matrix, iMessage, LINE and Nostr, plus the WebChat. On prediction markets, 10 platforms are listed, among them Polymarket, Kalshi, Betfair, Manifold, Metaculus and PredictIt, with the README claiming particular depth in Polymarket BTC/ETH/SOL/XRP binary markets across several round lengths with automated strategies. On perpetual futures, 7 exchanges are listed, including Binance, Bybit and Hyperliquid, with up to 200x margin.

The on-chain side is equally broad: full Solana integration (Jupiter, Pump.fun, Raydium, Orca) and several EVM chains via Uniswap V3 and 1inch, plus a Solana-native perpetuals protocol and Bittensor TAO mining. This breadth is the project's headline and also its main hazard: each venue is a separate set of credentials, behaviours and risks, and wiring many of them multiplies both the setup surface and the ways a strategy can go wrong. Enabling venues one at a time, and only after testing, is the prudent path.

Deploying with Docker and keeping it secure

For a persistent deployment, the repository ships a Dockerfile and a docker-compose.yml. The compose service builds the image, maps port 18789, reads secrets from an .env file, and persists a data volume for state and workspace, restarting unless stopped. The Docker image is based on Node 22 and pre-downloads a small embedding model at build time so the first request does not hang, which is a thoughtful touch for a service you leave running.

Security deserves attention because of what the credentials unlock. The .env holds your Anthropic key, messaging tokens, and, if you add them, exchange API keys and even wallet private keys such as POLY_PRIVATE_KEY. Those are the keys to real funds. Keep the .env off any shared machine, never commit it, run clodds secure to harden the setup, and prefer venues that use scoped API keys over ones that need a raw private key. Because the bot can act while you sleep, as the project's own description says, the blast radius of a leaked credential or a bad strategy is larger than for a tool you drive interactively.

Maturity, licence and the honest bottom line

Clodds is MIT licensed and moving fast: package.json is at version 1.9.1, recent releases are v1.9.1 on 2026-09-12 and v1.9.0 on 2026-08-31, and the last push was on 2026-09-12. It was built in 12 days for the Colosseum Agent Hackathon on Solana, which the README states plainly. The repository includes an AUDIT.md, a SECURITY.md, a test suite and a CI script, and there is an associated project token, which is a marketing element rather than a reason to use the software.

The honest bottom line: Clodds is an impressively broad, genuinely open-source piece of engineering, but it is a trading agent, and the interesting parts, autonomy, high margin and many venues, are exactly the parts that can lose money fast. Its hackathon origin and rapid 1.9.x cadence mean it is young and still changing. Treat it as a powerful tool for someone who already understands automated trading and its risks, start in read-only exploration, size any real trades small, and remember that no amount of AI strategy removes market risk or your own responsibility for the outcome.

Editorial conclusion

Consider CloddsBot only if you understand automated trading and are willing to risk real money on an autonomous agent. Install the current GitHub release, run clodds onboard, and start with WebChat and read-only exploration before adding any exchange keys. Guard your ANTHROPIC_API_KEY and trading credentials, run clodds secure, and never commit more capital than you can lose.

Frequently asked questions

Do AI trading bots like CloddsBot actually work?

CloddsBot provides the machinery, over 100 strategies, market connections and autonomous execution, but the README makes no verified profit claims, and no bot removes market risk. Automated trading, especially with high margin up to 200x, can lose money quickly; treat results as unproven and your own responsibility.

Are AI trading bots like this legal?

It depends entirely on your jurisdiction and the venues you connect. Some prediction markets and leveraged crypto products are restricted or banned in certain countries. CloddsBot is self-hosted software; whether you may legally use it, and each connected exchange, is your responsibility to check.

Can I run CloddsBot without connecting any exchange?

Yes. Only ANTHROPIC_API_KEY is required to run, and trading credentials are optional and separate. You can install it, run clodds onboard, and explore via WebChat and live market data without wiring any exchange keys, which is the safe way to evaluate it.

Official sources

  1. alsk1992/CloddsBot on GitHub
  2. License: MIT
  3. Project website
  4. README
  5. Releases
Community notes

Community notes