MCP Server

Official Talonic MCP server. Lets AI agents extract structured, schema-validated data from any document via the Model Context Protocol. Nine tools and two resources, verified end-to-end against production.

Hosted at mcp.talonic.com/mcp (the bare origin mcp.talonic.com also works) — zero install, works with every MCP client. Also available on npm as @talonic/mcp for local use, and listed on the official MCP Registry as io.github.talonicdev/talonic-mcp.

The server exposes nine tools: talonic_extract for document extraction with schema-validated JSON output, talonic_search for full-text search across your document registry, talonic_filter for structured field-level filtering, talonic_get_document for retrieving individual documents, talonic_to_markdown for converting documents to clean markdown, talonic_list_schemas for browsing available schemas, talonic_save_schema for persisting new extraction schemas, talonic_get_balance for checking the workspace credit balance and runway, and talonic_request_upload for browser-handoff uploads when the file is too large to send through the tool-call channel. All tools return typed, structured responses with confidence scores and provenance metadata.

Supported MCP clients include Claude Desktop, Cursor, Cline, Continue, Cowork, and any client implementing the Model Context Protocol specification. Setup takes one line — add the server URL to your client configuration and provide your Talonic API key as an environment variable.