Backtesting Arena MCP server

Backtesting Arena — Agent Skill

Battle-tested market intelligence for AI agents — every claim backtested, versioned, and explainable. Evidence and base rates, never financial advice.

Skill package for AI agents and LLMs that need programmatic access to the Backtesting Arena API: daily Bitcoin / crypto cycle scoring, on-chain indicators, aggregated strategy insights, and on-demand backtests.

Three channels, one product:

The full capability descriptor an agent should load is SKILL.md.

Install for Claude Code / Claude Desktop

# Coming soon when /install-skill GA's; for now clone + reference manually:
git clone https://github.com/Schoasch/skill-backtesting-arena.git

Reference skills/backtesting-arena/SKILL.md as a context document in your agent runtime, or use the MCP server directly (see below).

Quick start

1. Cycle-aware trading bot (Bearer)

curl https://tradingstrategies.work/api/arena-pulse/today
# → { "score": 35, "band": "risk-off", ... } — no API key required

Want higher quotas + full surface? Get a free Bearer key.

2. Autonomous agent with USDC wallet (x402)

# First call returns HTTP 402 + payment instructions:
curl -i https://tradingstrategies.work/api/v1/agent/btc-cycle/latest
# → 402 Payment Required, body has accepts[] with payTo, asset, amount, network

# Discovery descriptor lists all paid endpoints + pricing:
curl https://tradingstrategies.work/.well-known/x402

Full Python client showing the EIP-3009 sign + retry pattern: examples/python-x402-client.py.

3. MCP server (Claude Desktop, Cursor, etc.)

Two transports — HTTP (recommended, requires Bearer key) or stdio.

HTTP variant (drop into claudedesktopconfig.json → mcpServers):

{
  "backtesting-arena": {
    "url": "https://tradingstrategies.work/api/mcp",
    "headers": { "Authorization": "Bearer sk-arena-…" }
  }
}

Full example incl. stdio variant + Cursor config: examples/claude-desktop-config.json.

Examples folder

Discovery files (machine-readable)

For AI tooling that auto-discovers APIs:

• /openapi.json — full OpenAPI 3.1 spec, ~70 paths, x-x402 extension on agent endpoints
• /skill.md — this skill (canonical, kept in sync with this repo)
• /llms.txt — LLM-friendly summary
• /.well-known/api.json — cross-tool discovery hub
• /.well-known/x402 — Bazaar-v2 descriptor of paid endpoints (alias .json)
• /.well-known/ai-plugin.json — OpenAI-style plugin manifest

Pricing

Shortened. The full README is on GitHub.