Bird CMS documentation
Bird CMS is a markdown CMS your AI agent edits directly. Content is plain .md files on disk - no database on the hot path, no build step, no plugins. Pages are live the moment their files exist, and every page ships Schema.org markup, a sitemap and llms.txt by default, so it is built to be cited by answer engines.
Written for two operators
You do not have to be a developer to run a Bird CMS site. The whole system is designed around two operators working together:
The non-technical human →
You never touch code or files. You tell your AI assistant - Claude, Cursor, ChatGPT - what to publish, and it writes the markdown for you. Plain-language prompt recipes for publishing, editing, FAQs and rollback.
The AI agent →
The agent edits the site through a native MCP server and a REST API: it reads, writes, publishes and searches content directly. The content contract (meta.yaml schema) tells it exactly what valid content looks like.
Why Bird CMS
- Agent-native. A stdio MCP server exposes the site to Claude, Cursor, Continue and Zed - no API key to provision, no plugin, no copy-paste loop.
- Files on disk.
git diffis your audit trail,cp -ris your backup,grepis your search. No database to run or restore. - AEO from day one. Schema.org for 13+ types, sitemap, robots and
llms.txtare generated automatically. - Featherweight. The engine is a couple of megabytes - small enough to read end-to-end - with atomic versioned-symlink deploys and one-command rollback.
Start here
Quickstart →
From clone to a published page in a few minutes.
Content model →
How articles and pages live on disk: markdown + meta.yaml.
New here? Run through the Quickstart, then pick your track: non-coder or AI agent.