Document processing for agents

No. The hosted MCP server uses OAuth: your client opens a browser sign-in and you choose which workspaces it may use and whether it acts in Test or Production. API keys are for the CLI in headless environments and for SDK code.

MCP is the best fit for chat and tool-calling clients. For agents writing durable integration code, use an SDK with the extend-api skill; for one-off shell-driven jobs, the CLI with the extend-cli skill. The paths are independent, so set up only what your agent will use.

Chat clients don't expose local files to remote servers. When you ask for a file that's only on your machine, the agent requests an upload link: open it, sign in to the Extend dashboard, drop your files, and click Done. Links expire after 15 minutes and only members of the target workspace can use them. Documents at a public URL need no upload.

Every operation targets a specific workspace and environment, and the agent must state both on every call. On the OAuth consent screen, grant the client the Test Environment only. Resources are separate between Test and Production, so an extractor created in Test never exists in Production.

Append a comma-separated list of group or tool names in the tools query parameter, for example https://mcp.extend.ai/mcp?tools=extract,files. The groups are extract, classify, parse, split, workflows, edit, files, webhooks, and evaluations. Access and documentation tools stay available.

Yes. https://docs.extend.ai/llms.txt indexes every page, and appending .md to any docs URL returns clean Markdown. Every page on extend.ai has a Markdown twin too. Agents on the MCP path get the same through the search_documentation and get_documentation tools.