Connect AlpaTrade to your AI client.
A public Model Context Protocol server for AlpaTrade strategies and agent discovery. Connect once, then ask your client to search strategies, inspect performance and read published strategy skills.
https://alpatrade.chat/mcp Streamable HTTP · no sign-in · read-only public data
Claude Code · Cursor · any client that supports MCP over streamable HTTP
Connect your client.
Run this command in your terminal:
claude mcp add --transport http alpatrade https://alpatrade.chat/mcpAdd this server in Cursor Settings → MCP, or save it in ~/.cursor/mcp.json:
{
"mcpServers": {
"alpatrade": { "url": "https://alpatrade.chat/mcp" }
}
}Use the URL in any streamable HTTP MCP client. To inspect tools with curl:
curl -sS https://alpatrade.chat/mcp \
-H 'Accept: application/json, text/event-stream' \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'What the MCP tools can read.
list_public_strategies Search public strategies with live or clearly marked backtest figures.
get_public_strategy Read one public strategy's return, period, author and source link.
get_public_strategy_skill Fetch the public Markdown strategy skill for your agent.
list_agent_interfaces Discover the agent catalogue and its authenticated API endpoints.
Results link to their public AlpaTrade source. Annualised return is the simple period return × 252 / trading days; backtests are hypothetical. Private account data and trading actions require the authenticated API below.
Build on the AlpaTrade agent desk.
Call the primary LangChain DeepAgent, research analysts, backtester, validator, paper trader, reconciler, reporter, and durable autonomy pipeline through typed APIs.
Backtest from your terminal.
The alpatrade CLI runs the same backtester as the web desk and writes every run to a dated folder you can diff, re-run and share.
Example output — illustrative figures, not a live result.
Command name format
btd-7dp-05sl-1tp-1d-3m reads as: btd buy-the-dip strategy · 7dp 7% dip threshold · 05sl 0.5% stop-loss · 1tp 1% take-profit · 1d 1-day hold · 3m 3-month lookback. Pattern: {strategy}-{param}-…-{lookback}; 0.5% is written 05.
The artifacts folder
Each run lands in backtest-results/<date>_<symbol>_<strategy>_<interval>/: config.json and strategy_spec.json (parameters), trades.csv and round_trips.csv, equity.csv with a benchmark curve, summary.json (trades, win-rate, Sharpe, max drawdown), data_fingerprint.json and fee_source.json, report.md / notes.md, and a run.py that reproduces the run.
Choose the right specialist for the job.
Every agent below is callable through a typed HTTP contract. Skills describe stable capabilities; follow the reference link for request bodies, response schemas, examples, authentication, and error responses.
DeepAgent interfaces
3 agentsDeepAgent Assistant
POSTCanonical durable, tenant-safe DeepAgents API with JSON or SSE responses.
DeepAgent Assistant (JSON compatibility)
POSTCompatibility JSON chat interface backed by the shared DeepAgents service.
DeepAgent Assistant (SSE)
POSTCompatibility SSE chat; anonymous callers receive public research only.
Research agents
4 agentsPremarket Agent
POSTRead-only premarket scan and ranked mover intelligence.
Alpha Growth Agent
POSTEvidence-backed growth-company research report.
Alpha Value Agent
POSTEvidence-backed value-company research report.
Alpha Comparison Agent
POSTGrowth and value views generated from one shared evidence collection.
Analysis and validation
4 agentsBacktest Agent
POSTRun the parameter-grid backtest and return the best configuration.
Validation Agent
POSTValidate stored backtest or paper trades against market data.
Report Agent
GETReturn run summaries, detail, P&L, and ranked strategies.
Daily Trading Advisor
GETRead persisted, evidence-gated paper strategy and risk reviews.
Paper trading operations
2 agentsPaper Trade Agent
POSTStart a durable paper-trading subprocess and return immediately.
Reconciliation Agent
POSTCompare the position ledger with the linked Alpaca paper account.
Agent orchestration
2 agentsFive-Agent Orchestrator
POSTRun backtest, validation, paper trading, and reporting as one workflow.
Autonomy Scout
POSTScan candidates and enqueue the durable paper-only agent pipeline.
Human guides and machine contracts.
Use ReDoc for reading and Swagger UI for interactive calls. OpenAPI and catalogue JSON remain machine-readable for SDK generators, agents, and service discovery.
Machine-readable: OpenAPI JSON · Agent catalogue JSON
User JWTs and service API keys.
User clients send an Authorization bearer token obtained from /auth/login. Trusted services send X-API-Key and include X-User-Id when acting for a user. All trading and autonomy routes are paper-only.