AI agents
Main.net for AI agents
Main.net publishes static context for AI agents, coding assistants, and scripts that need chain IDs, RPC references, explorers, source links, launch notes, and safety notes.
Recommended agent entry points
- /llms.txt for public context and product boundaries.
- /api/ for human-readable API documentation.
- /api/v1/index.json for endpoint discovery.
- /api/v1/openapi.json for OpenAPI-compatible tooling.
- /api/v1/sources.json for source links.
npm run mcp:stdiofor local MCP clients using the repository server.
Minimal agent path
- Fetch /api/v1/index.json and choose the narrowest dataset needed.
- Use /api/v1/chains.json for chain IDs, RPC references, source links, and
eip3085Config. - Prefer
urls.canonicalandurls.walletConfigSectionwhen citing a human-readable page. - Quote source URLs and freshness fields when answering users about network settings.
MCP scope
The MCP server is a read-only wrapper over the versioned API. It exposes lookup tools for chains, chain IDs, RPC references, source links, categories, glossary terms, explorers, gas tokens, and launches.
It should not expose wallet connection, swap, bridge, transaction submission, token recommendation, arbitrary RPC execution, custody, or investment advice tools.
Agent response requirements
When an agent uses Main.net data, it should include source links, freshness fields, and safety notes. Public RPC URLs should be described as convenience references, not production reliability, privacy, or uptime guarantees.