Three parts: a generated Soul.md, an MCP server, and a nine-phase conversation engine.
Daemon is a personal intent engine. It generates a Soul.md, serves an MCP server, and runs its conversation in nine phases.
Its own repo puts the idea in one line: an AI that interviews you deeply, then makes every other AI tool know you instantly.
One file, one server, one conversation.
Soul.md
Generated by the engine — the artifact the rest of it produces, and the thing at the centre of the hero above.
MCP server
The engine exposed as tools a model client can call, rather than an app you have to visit.
Nine-phase conversation
The engine runs in nine phases, in order. Nine is the file's number, not a flourish — the hero is that loop.
One version out, and a move to a new host under way.
Built on Next.js and Neon Postgres. v0.1 was live on Vercel and is offline while it moves to Cloudflare Workers, a move that is written and not yet deployed. It comes back with its AI off at first: the Claude routes answer 503 until a key is set.
Whether the database v0.1 used still exists, and whether Creator Codex needs any of it.