For agents and developers

Every machine-readable surface of devrel.directory in one place - Markdown mirrors, the MCP server, feeds, the crawler policy, and rate limits.

devrel.directory is built to be read by machines as much as by people. Every surface below is read-only, unauthenticated, and free to use: no API key, no signup, nothing to negotiate. This page is the index; where a surface has deeper documentation, it links there.

Markdown mirrors

The site's content is available as plain Markdown, following the llms.txt convention:

/llms.txt: the index. The site's name, a one-line summary, and a linked map of the docs, blog, directory, jobs, events, feeds, and the MCP server.
/llms-full.txt: the full corpus. Every docs page plus every directory listing, active job, and upcoming event, in one Markdown file.
Markdown twins: append .md to a page URL to get that page as raw Markdown. This works for every docs page (/docs/<path>.md, including this one) and every directory listing (/directory/<slug>.md).

The thinking behind this setup is covered in Writing Docs LLMs Can Use; this site is its own reference implementation.

MCP server

A remote Model Context Protocol server runs at https://devrel.directory/mcp, speaking Streamable HTTP. It is read-only and unauthenticated, and exposes six tools: search_directory, get_listing, list_jobs, list_events, directory_stats, and search_docs. It is published to the official MCP registry as directory.devrel/devrel-directory. Connection snippets and full tool descriptions are on the MCP server page.

Feeds

/rss.xml: blog posts.
/events.ics: every upcoming DevRel event as an all-day iCal entry.
/cfps.ics: open CFP deadlines as timed iCal entries, each with a one-week reminder.

Crawler policy

AI crawlers are explicitly welcome. robots.txt names GPTBot, ClaudeBot, Claude-User, PerplexityBot, and Google-Extended with allow-all groups of their own, and the generic group allows everything for every other crawler too. The only disallowed paths, for all user agents, are the payment and webhook endpoints under /api, which render nothing worth indexing. The sitemap linked from robots.txt covers the site's indexable pages.

Rate limits and fair use

Two surfaces are rate limited per client, because each request does real work:

/mcp: 60 requests per minute.
/llms-full.txt: 30 requests per minute.

Requests over a limit get a 429 with a Retry-After header; back off for that long and retry. Everything else is currently unthrottled, subject to fair use. If you plan sustained high-volume access, email [email protected] first so it can be accommodated rather than mistaken for abuse.

Attribution

If your agent or product cites data from here, please link back to devrel.directory so readers can find the source.