MCP
Archive's Model Context Protocol server lets AI agents work with your data directly. It exposes the same surface as the GraphQL API: 64 tools, each wrapping exactly one public root field. For field-level detail on what a tool returns, follow it through to the API reference.
01 Overview The Archive MCP server exposes the same /api/v2 surface as agent tools over Streamable HTTP, with one tool per public GraphQL root field. 02 Use with Claude Connect Archive to Claude.ai, Claude Desktop, or Claude Code with the endpoint URL alone, authorize over OAuth 2.1, and verify the connection. 03 Connect a client Point an MCP client at Archive with a bearer token or the OAuth 2.1 flow, verify the connection with getWorkspaces, and understand what tools/list returns. 04 Agent conventions Rules that apply across every Archive MCP tool — userErrors versus isError, retry discipline, destructive tools, async ingest, concise responses, and view filters. 05 Example prompts Working prompts for the Archive MCP server — content ranking, creator lookup, saving a view, importing a post, and a structured weekly-report workflow. 06 Read tools The 33 tools that only fetch data, each wrapping one public query field. 07 Write tools The 31 tools that change something, each wrapping one public mutation. 08 Troubleshooting Common Archive MCP problems and their fixes — 401s, missing tools, unresolved workspaces, empty result sets, rejected writes, throttles, and client-specific issues. 09 Security & data handling How the Archive MCP server authenticates requests, isolates workspaces, audits writes, and what data a connected agent can reach. 10 Support & contact How to reach Archive about the MCP server, what to include in a report, and what to check first.