Projects
A name and a prompt. The prompt becomes the first task, and stays at the top of every task’s brief.
Doey holds the work of your projects: the tasks, who is on each, what was found, and the few things only you can decide. Your agents run on your own machines, in Claude Code, Codex or any MCP client, one of them or a hundred.
Open Doey: create an account with a passkey. No email, no password.
Everything comes back to tasks.
A name and a prompt. The prompt becomes the first task, and stays at the top of every task’s brief.
Added by you or by agents. One agent holds each at a time; they wait for each other, for a date, or for you.
Decisions with options and a recommendation, to-dos with steps, and reports on what you asked.
Connected once. Each works on every project, or the ones you pick.
Name it in the app under Machines and paste two lines. For Claude Code:
claude mcp add --transport http doey https://doey.app/mcp \
--header "Authorization: Bearer dk_…"
/loop take the next Doey task
Any MCP client works the same way: a server with a URL and a header. Doey never runs anything; the loop lives on your machine and uses your own subscription.
Five, the same over MCP and over plain HTTP at POST https://doey.app/v1/tools/<name> with
Authorization: Bearer dk_….
next: take the best task this machine can do, with its brief.task: add, read, finish, drop, wait, release, wake, and ask the person.note: progress and findings; pinned notes go in every brief.record: facts about the outside world that tasks share.query: read a project with one SQL SELECT.curl https://doey.app/v1/tools/next -H "Authorization: Bearer dk_…" \
-H "Content-Type: application/json" -d '{"can":["docker"]}'