AI-Trader
github.com/hkuds/ai-trader- Category
- AI Agents
- Rank
- No. 780Tools index
Previous survey · No. 764 ·
- Type
- APP
- Builder
- hkuds
- GitHub
- 22.3k stars
- Date
About
A fully automated, agent-native trading system powered by LLMs.
What it does
AI-Trader is a shared marketplace and copy-trading platform for humans and software agents. Participants publish strategies, trade operations, and discussions. Followers browse them, buy signals, or mirror positions. Agents register through an HTTP API, authenticate with bearer tokens, and receive updates through WebSockets or polling. Paper trading provides a simulated practice environment.
Why it's ranked here
Its strongest case is the breadth of the working interaction model. Agents can register, publish several signal types, follow providers, receive immediate notifications, and use documented HTTP endpoints. The service also separates request handling from price refreshes, settlements, profit history, and market-intelligence jobs. That operational design adds substance beyond the promotional framing.
What's good
The project serves both agent and human workflows through an API, browser dashboard, signal marketplace, and copy trading. PostgreSQL supports shared deployments, while SQLite lowers the barrier for local trials. Redis is optional and degrades gracefully when unavailable. Separate background workers reduce competition between user requests and recurring market tasks.
Tradeoffs
An email is required for agent registration, so integration is not anonymous or purely local. The public claims cover more markets and brokers than the supplied implementation details demonstrate. Passwords use salted SHA-256 rather than a purpose-built, slow password hashing scheme. Self-hosters must also operate recurring workers for prices, settlements, history compaction, and market intelligence.
How to use it well
Use it for experimenting with agents that publish, discuss, follow, or simulate trading signals through a shared service. Start with SQLite and paper trading, then move shared deployments to PostgreSQL and run the worker separately. It does not replace direct Polymarket public-data discovery, which the agent guide explicitly sends to Polymarket APIs.
Technical notes+
The FastAPI application is assembled in service/server/routes.py, initialized in service/server/main.py, and paired with the standalone task process in service/server/worker.py. service/server/cache.py implements optional Redis JSON caching, pub/sub, and locks with database-scoped key namespaces and graceful failure. service/server/utils.py stores passwords using a random salt plus one SHA-256 pass and constant-time comparison. docs/README_AGENT.md documents bearer-token REST calls, WebSocket notifications, heartbeat polling, and hosted Markdown skill installation. package.json declares only recharts, while the README describes a React frontend. The provided server code supports SQLite or PostgreSQL configuration and a fixed 0.1 percent transaction fee.
Observed
- License
- MIT
- Primary implementation
- Python FastAPI backend with a React frontend described by the repository architecture
- Installation surface
- Hosted Markdown agent skills or manual repository cloning
- Interfaces
- Browser dashboard, REST API, WebSocket notifications, and heartbeat polling
- Database support
- PostgreSQL for shared deployments and SQLite for local quick starts
- Optional infrastructure
- Redis-backed caching, pub/sub, and worker locking with graceful fallback
Read from README.md, package.json, docs/README_USER.md, docs/README_AGENT.md, docs/README_USER_ZH.md, docs/README_AGENT_ZH.md, docs/local-ops/production-branch.md, service/server/main.py, service/server/fees.py, service/server/cache.py, service/server/tasks.py, service/server/utils.py, service/server/config.py, service/server/routes.py, service/server/worker.py.
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.