Cykl in Claude and ChatGPT

1. Connect in Claude

On claude.ai or in the Claude app: Settings → Connectors → Browse connectors → Cykl → Connect, once Cykl is listed there.

Until then, or if you prefer, add it yourself: Settings → Connectors → Add custom connector, with this URL:

https://app.cykl.io/mcp

In Claude Code, run:

claude mcp add --transport http cykl https://app.cykl.io/mcp

2. Connect in ChatGPT

Settings → Plugins → Browse directory → Cykl, once Cykl is listed there.

3. Signing in and disconnecting

Connecting opens Cykl: you sign in with your Cykl account and approve the connection. Your AI never sees your password.

You can disconnect at any time in Cykl, under Settings → Integrations. The connection stops at once.

4. When changes land

Cykl keeps your plan on your own devices, so a change your AI makes is queued and applied the next time your Cykl app is open — the same way a tap would apply it. Until then your AI sees it as pending.

Every change an AI makes is listed in Settings → Integrations, and each one can be undone there.

5. Things to ask

• “What did I commit to this quarter, and how is my week going against it?”
• “Add a project to write the Q4 review, and put its first step on Friday.”
• “Summarize my last four weekly reviews.”

6. Every tool

Reading (9)

• Current state get_current_state — the open cycle at every level, today’s actions and habits, active projects and your mastery score.
• List projects list_projects — your projects, filtered by status, timeframe, domain or priority.
• Get a project get_project — one project in full: its definition of done, sub-projects, cycles, open actions and summary.
• List cycles list_cycles — your cycle history, from days to a lifetime, most recent first.
• Get a cycle get_cycle — one cycle in full: focus, metrics, habits or routines, review, and the cycles above and below it.
• List actions list_actions — your actions: a day’s plate, a date range, or the undated Next Actions inbox.
• Get reviews and lessons get_reflections — what went well, what didn’t and what you learned, cycle by cycle.
• Check queued changes get_change_status — the changes your AI asked for, and whether each is pending, applied or refused.
• Search search_state — find a project, action or cycle by text.

Changing (10)

• Open a project create_project — a new project with a definition of done, at a timeframe, optionally under a parent.
• Change a project update_project — its name, status, priority, definition of done, parent or domains; only what is sent changes.
• Tick criteria of a project tick_criteria — mark lines of its definition of done as met or not, which moves its progress.
• Write a project’s summary write_project_summary — replace the few lines a project is remembered by.
• Add an action create_action — an action on a day’s plate, in the Next Actions inbox, or logged as already done.
• Complete an action complete_action — mark an action and its sub-steps done.
• Add bullets to a cycle append_cycle_bullets — add lines to a cycle’s focus or review; never rewrites yours.
• Set what a chapter, season or lifetime distils set_higher_cycle — manifesto, values, identities, skills, vision outcomes and anti-vision.
• Save a vision exercise save_vision_exercise — the result of a self-discovery or vision exercise on a chapter, season or lifetime.
• Record readings from a device record_readings — sleep, weight, heart rate or steps from a watch, ring or scale; a value you typed always wins.

7. Privacy and support

A connected AI receives only what the tools it calls return. What the AI provider does with it is governed by its own terms. The details are in our privacy policy.

Something not working? Write to [email protected].