CIP — Capital Investment Prospects
CIP (Capital Investment Prospects) turns natural language into backtested,
risk-managed algorithmic trading strategies.
Paper-trading sandbox and research tool — not a broker, not investment
advice. See docs/exec-summary.md for the full product
vision and docs/architecture.md for how the
four-agent pipeline fits together.
Status
The strategy JSON schema (schemas/strategy.json),
its validator layer (src/cip/validation.py), and a
stub LangGraph four-agent pipeline (src/cip/agents/) run
end to end under test, wrapped by a FastAPI layer (api/) with SQLite
persistence and SSE, and a Next.js frontend (frontend/) with
symbol search and per-asset technical analysis.
Real: the schema and validators, the Risk Cop's hardcoded gates and
correction loop, and live crypto market data from Binance. Stub: the four
agents' reasoning (fixture-based selection, deterministic pseudo-metrics, no
LLM calls), the backtest numbers, and equities/forex candles until
POLYGON_API_KEY is set — anything stubbed is badged as such in the UI.
Screener, variant lab, decision memos, and the monitor are not built yet;
see docs/decisions/0002-web-architecture.md.
To run what exists:
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest # 33 tests, no keys or network needed
uvicorn api.main:app --port 8000 # API
npm install --prefix frontend && npm run dev --prefix frontend # UI on :3000
Docs
Contribute
Branch rules and how to run locally: CONTRIBUTING.md.
If you're an AI agent working in this repo, read CLAUDE.md
first — it encodes hard constraints (no live execution, no raw codegen,
backtest integrity rules) that override anything implied here.
