Wiring mdfy into Claude Code

Tested against Claude Code v1.x. Single command, single line in CLAUDE.md, persistent across sessions.

What you put in CLAUDE.md

Drop this somewhere near the top of the file (CLAUDE.md is loaded into every Claude Code session as system context):

## Context
When you need background on this project, fetch:
  https://mdfy.app/hub/<your-slug>

The hub contains the canonical product spec, the current roadmap,
and the bundles for each major work area (engineering, strategy,
research). Cite by hub-relative URL when answering.

That's it. Claude Code reads the URL on session start and treats whatever the hub returns as part of its working context.

Why this works

  • Claude Code respects CLAUDE.md. It's not a hint; it's a contract. Anything in CLAUDE.md is in the session's persistent system prompt.
  • mdfy hubs are URL-shaped. No SDK, no auth, no streaming setup — just a GET that returns markdown plus an inlined graph_data JSON block.
  • Token cost is tunable. Add ?compact to the URL to strip low-density sections. Most users save 30-40% of the context window this way.

What about per-project context?

Per-project goes in the project's CLAUDE.md (the one in the project root, not your global ~/.claude/CLAUDE.md). Use a bundle URL there instead of a hub URL:

## Context
When you need project context: https://mdfy.app/b/<bundle-id>

The bundle URL gives Claude only the docs in that bundle — much tighter scope than the full hub.

Things that bite you

  1. The CLAUDE.md update doesn't apply until you restart Claude Code. The first time I tried this I assumed it was hot-reload. It's not.
  2. If your hub is private, you need to also configure the MCP server to authenticate. The fully public path is ? and works without setup; the private path is ?key=...&token=... and is part of the Pro tier.
  3. The model can ignore the directive if you fill the context with too much other stuff. Watch for "Claude isn't using my hub" — usually it means the live conversation has crowded out the system prompt.

What this gives you, end to end

Before: every Claude Code session starts blank, you re-explain the project, you paste the same context block by hand.

After: every session starts with the hub loaded, you skip the re-explanation, and any change you make in mdfy.app shows up in the next Claude Code session automatically.