Connect Codex to Hatchlog
Open Connect agents in Hatchlog and click Connect next to Codex. Hatchlog adds itself to Codex as an MCP server. On its next session, Codex can read your board and move cards. Token cost is not measured for Codex — that is Claude Code only.
Connect in one click
Open Connect agents from the Hatchlog sidebar. Find Codex in the list and click Connect. Hatchlog adds itself to Codex as an MCP server and leaves your other tools alone. You don't edit anything by hand.
What "Connected" means
When the row reads Connected, the Hatchlog tools are available to Codex. On its
next session, Codex can ask for the next task with next_task, plan
work with write_plan and add_tasks, and
report back with update_task. Start a new session for the tools to
load.
Token cost is Claude Code only
Hatchlog shows what each task cost in tokens for Claude Code. It does not measure token cost for Codex. Codex still reads your board and moves cards the same way — only the cost figures are left blank for its work. See Token cost for what is measured.
Check it works
In the Connect agents sheet, click Test connection. Hatchlog confirms that Codex can reach your board. In Codex, you can also just ask it to list your projects — if it answers from Hatchlog, you're set.
Repair a connection
If Codex stops seeing Hatchlog — after an update, say — open Settings → Agents and choose Repair for Codex. Repair puts the connection back. You never set anything up by hand.
Common problems
How do I know Codex is connected?
The Connect agents sheet shows "Connected" next to Codex, and Codex can call the Hatchlog tools on its next session.
Does Hatchlog track token cost for Codex?
No. Token cost is measured for Claude Code only. Codex can still read and update your board.
Codex doesn't see Hatchlog. What now?
Open Settings → Agents and choose Repair for Codex. Repair puts the connection back. You never set anything up by hand.