Vibe-Trading
github.com/hkuds/vibe-trading- Category
- AI Agents
- Rank
- No. 702Tools index
Previous survey · No. 692 ·
- Type
- APP
- Builder
- hkuds
- GitHub
- 33.5k stars
- Latest release
- v0.1.15
- Date
About
A personal, agent-driven trading assistant.
What it does
Vibe-Trading turns natural-language finance questions into research, market analysis, backtests, valuations, and scheduled workflows. Users can work through an interactive terminal, web interface, REST service, or compatible model client. Broker connections exist, while the model-facing tool interface remains research-only and cannot place or cancel orders.
Why it's ranked here
The breadth is unusually practical: one package combines research, quantitative finance, backtesting, broker reads, automation, and several interfaces. Its strongest argument is safety-conscious engineering, including evidence checks, guarded network access, disabled shell tools by default, and explicit refusal of incomplete valuations. The large dependency and configuration surface makes it a serious system, not a lightweight helper.
What's good
Research results face concrete checks around instrument identity and observed price evidence. Missing valuation inputs stop a model instead of triggering silent defaults. The model-facing server exposes read-only finance and broker tools, with no order placement or cancellation. Network transports validate hosts and browser origins, while shell access requires explicit operator consent. Python and frontend test configurations are both present.
Tradeoffs
The base installation pulls in a broad scientific, document, web, server, and agent stack. Specialized brokers, statistics, markets, messaging channels, and OpenBB support require separate extras or local credentials. Live trading adds mandates, connector setup, authorization, halt controls, and runner state, so operation carries real complexity. Some free research markets need no data key, but paid routed research and broker features have their own requirements.
How to use it well
Use it when one workflow must connect conversational research, market data, finance calculations, backtests, scheduled studies, and read-only broker context. Start with the terminal or web interface, then expose the research tools to a compatible model client or build automation against the REST service. Treat live trading as a separately configured, supervised layer. It does not replace broker accounts, paid data access, or operational risk controls.
Technical notes+
pyproject.toml packages vibe-trading-ai for Python >=3.11,<3.14 and registers vibe-trading plus vibe-trading-mcp console scripts. agent/api_server.py assembles a FastAPI application covering runs, sessions, settings, uploads, channels, swarms, live trading, scheduled research, and optional OpenBB routes. agent/mcp_server.py uses FastMCP, supports stdio, legacy SSE, and Streamable HTTP, keeps shell tools opt-in, and wraps network transports with Host and Origin guards. frontend/vite.config.ts defines the React development proxy and production chunking; frontend/vitest.config.ts configures jsdom tests and V8 coverage. tools/ci_env_var_gate.py enforces centralized environment reads through AST scanning, with tests in tools/test_ci_env_var_gate.py.
Observed
- License
- MIT
- Primary language
- Python, with a React and TypeScript frontend
- Python support
- Python 3.11 through 3.13
- Packaging
- Published as the pip package vibe-trading-ai with optional feature extras
- Interfaces
- Interactive CLI, web UI, REST API, and MCP server
- MCP transports
- stdio, legacy SSE, and Streamable HTTP
- Testing
- Pytest is configured for agent/tests, and Vitest is configured for frontend tests
Read from README.md, pyproject.toml, wiki/main.js, agent/api_server.py, agent/mcp_server.py, frontend/vite.config.ts, tools/ci_env_var_gate.py, frontend/vitest.config.ts, frontend/postcss.config.js, frontend/tailwind.config.ts, tools/test_ci_env_var_gate.py, agent/cli/main.py, wiki/docs/main.js.
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.