Best for: Driving Traseq from code or an AI assistant

Use Traseq from code or an AI assistant

You write the strategy; Traseq checks it, saves it, runs the backtest, and returns the result. Strategies use one format everywhere — SignalGraph v2 — so what you build in code matches what the no-code editor produces.

Key concepts

API key: A credential your code uses to prove it belongs to your workspace. It is created in workspace settings, and each plan allows a set number at once.
Capabilities: A live list of what your workspace supports right now — which indicators, operators, and markets. Reading it first means your code never hardcodes assumptions that could break later.
Validation: Every strategy you send is checked for both format and logic. A rejected one comes back with the exact field and the fix needed, so an AI assistant can correct itself without a human.
Draft and finalize: A strategy is saved as a draft first, then locked into a fixed version. Backtests run only against a locked version, so a result stays reproducible.
Queued backtest: Backtests run in the background. You submit one, receive a run id, then check back for the result instead of waiting on the request.

How this looks in Traseq

An API key is created in workspace settings.
A call to /capabilities confirms the workspace supports what the strategy needs.
A SignalGraph v2 strategy is written, validated, then finalized into a version.
A backtest is queued against that finalized version, and the run is read back.

Install and use

One npm package connects Traseq to your AI assistant. It needs Node.js 20 or newer and a workspace API key (trsq_…). It never places live orders — your agent reasons and writes the strategy; Traseq handles validation, backtests, and results.

@traseq/agentFor AI assistants

An MCP server, CLI, and high-level research SDK. It lets Claude Code, Cursor, and similar assistants author, validate, and backtest strategies in your workspace. The setup wizard stores your API key in the OS keychain — no plaintext secret in any config file.

bash

npx -y --package @traseq/agent@latest traseq-agent setup

Ask your connected AI assistant to start

Use the Traseq MCP server. First call start_research_engagement with the prompt
"Validate a BTCUSDT 4h trend-following idea." If the tool is unavailable, tell me
the Traseq MCP server is not connected.

Your next step