Backtesting Arena

What backtesting-arena does

The backtesting-arena MCP server connects AI agents to the Backtesting Arena API at tradingstrategies.work. Its subject matter is Bitcoin and crypto market research, with capabilities covering daily cycle scoring, on-chain indicators, aggregated strategy insights, and backtests requested on demand.

The package is intended for agents that need reproducible market analysis rather than unqualified trading recommendations. The project describes its outputs as backtested, versioned, and explainable, and emphasizes evidence and base rates. It also states that the service is educational and research-oriented, not financial advice or a signal-selling product.

How it works

The recommended MCP transport is HTTP. A client connects to https://tradingstrategies.work/api/mcp and sends a Bearer API key in the Authorization header. The repository also documents a stdio configuration, but the excerpt does not provide a runnable local server command for that mode.

The underlying API exposes more than the MCP endpoint. Public endpoints can be queried without credentials and use a fair-use cache, while the REST and MCP surface uses an API key for broader access and higher quotas. The API also offers x402 pay-per-call access using an EIP-3009 wallet signature and USDC on Base; that payment mechanism is described for autonomous API clients rather than as the authentication method for the documented MCP HTTP configuration.

The repository includes a TypeScript MCP client example that lists 87 tools and calls one of them. Discovery resources include an OpenAPI specification, an agent skill document, an LLM-oriented summary, and API manifests hosted by the service.

Setup and configuration

For Claude Desktop or Cursor, configure the remote MCP URL and provide a Bearer credential, following the repository's example configuration. A free Bearer key is available through the Backtesting Arena account dashboard. The README says that the repository can also be cloned and that skills/backtesting-arena/SKILL.md can be referenced as a context document, but it does not describe that clone as a standalone MCP server installation.

The available pricing options affect API access. Public endpoints and the Free Bearer tier are available at no charge, with the stated free limits of 30 requests per hour and 200 per day. Paid API tiers add higher quotas and access to features such as backtest triggers and larger or broader backtest windows. The service also lists per-call x402 prices from $0.01 to $0.50 USDC.

Tools and capabilities

Depending on the access channel and plan, the backtesting-arena MCP server can support:

• Reading the daily Bitcoin or crypto cycle score.
• Retrieving on-chain market indicators.
• Querying aggregated strategy insights.
• Triggering on-demand backtests.
• Working with larger windows or multi-pair universe backtests on qualifying paid plans.
• Accessing machine-readable API and skill descriptors for agent discovery.

The README does not enumerate each MCP tool or define every request parameter, so implementers should consult the linked OpenAPI and skill documents for the current surface.

Limitations and notes

The MCP HTTP configuration requires a Bearer API key, and quotas depend on the selected tier. Public access is narrower and subject to fair-use caching. The API itself is governed by its service terms, while the repository's skill package is licensed under MIT.

Results should be treated as research outputs. The project explicitly avoids presenting them as financial advice. API pricing is reviewed quarterly, and the service was described as public since 2026, so clients should use the live discovery documents rather than assuming the repository examples define the complete or permanent API.

Getting started with this backtesting-arena MCP server

Always refer to the official documentation for the most accurate and up-to-date information.