A local financial terminal your AI agent operates for you. Market data, backtests, paper books, news — driven in plain language, running entirely on your machine.
Needs uv and Python 3.12. No clone, no build, no server to start:
claude mcp add otto -- uvx --from git+https://github.com/0Smallcat0/otto otto-mcpThen ask for things:
- "Refresh market data and show me the BTC snapshot."
- "Backtest an SMA cross on BTCUSDT and tell me if it's any good."
- "What do you make of my holdings?"
Want the dashboard as well: npm --prefix frontend install && npm --prefix frontend run build,
then uv run otto → http://127.0.0.1:8765/
It keeps score. The agent records dated calls — a stance, the reasoning, the price level that would prove it wrong, a horizon — and every call is later graded against the price the market actually printed. Calls without reasoning are refused rather than guessed, moves inside a flat band don't count as skill, and a call graded long after it matured is excluded because it measured a window its thesis never claimed. The hit rate is whatever the market says it is.
"An AI can operate it" is a benchmark here, not a tagline. A real headless agent gets 20 plain-language tasks in hermetic sandboxes, graded programmatically on terminal state and artifacts — never by an LLM judge:
| Model | Passed | Avg turns |
|---|---|---|
claude-sonnet-5 |
20 / 20 | 6.5 |
claude-haiku-4-5 |
19 / 20 | 6.8 |
Safety tasks grade refusal: asking for a live order must leave state unchanged. Live trading, credential entry and code execution aren't switched off — they're unreachable through the surface the agent has.
One typed contract (137 actions across 16 routes) is the single source of truth; the MCP tools, the UI capability catalog and the eval suite are all derived from it. 680 tests on Windows + Linux CI.
- Architecture · ADRs
- AI operator guide
- Eval methodology and full results
- Clean-room boundary — built by observing a reference terminal's workflow, never by reading or porting its code; a test fails the build if that wall is crossed.
python -m pytest -q && python -m ruff check .MIT.