Hamloy for Claude

A chat with Claude is a good place to think a project through. It's a poor place to keep the result. The plan stays in that conversation; the agent that builds it in a terminal tomorrow never sees it, and the questions it runs into reach you nowhere.

With Hamloy connected, Claude writes the plan into your project's lists as you talk: tasks with subtasks, what waits on what, the decisions that are yours as questions with options, and what only you can do as requests with the steps. Claude Code or Codex picks it up from there, and you see it all on one board.

Plan in the chat

Ask Claude to put the plan in Hamloy, and it calls the tools. Claude shows each call in the chat; here are two, with Hamloy's replies word for word:

add({"kind": "question", "title": "Where should synced habits live?",
     "options": ["Supabase", "Our own Postgres on Fly.io", "iCloud only, no server"], "recommended": 1})
Q-sy2c@1  Where should synced habits live?  options: 1) Supabase (recommended)  2) Our own Postgres on Fly.io  3) iCloud only, no server

add({"title": "Offline sync with a local SQLite cache", "section": "Sync", "needs": ["Q-sy2c"]})
B-se4t@1  Offline sync with a local SQLite cache  (waits on Q-sy2c)

The question is yours now. It waits on your board, with Claude's recommendation marked, and the task that needs it waits too.

Build elsewhere

The same lists reach every agent you connect. Open Claude Code in your repo the next morning, and with the hamloy program its session starts knowing what's open and which task to take next. It claims the task, so a second session or Codex picks something else, and ticks it when the work is committed. See Hamloy for Claude Code and Hamloy for Codex.

Answer from your phone

An agent that needs a decision doesn't wait at a terminal for you. It asks, and carries on with the work every answer needs:

You can answer in two places. Open your board on your phone and pick an option. Or ask Claude in the phone app, "what's waiting for me?", and tell it your answer. Claude records it, and every agent waiting on it hears it with its next tool reply:

answer({"key": "Q-sy2c", "answer": "1", "why": "Auth is built in"})
Q-sy2c@2 answered: Supabase
no longer waiting on it: B-se4t

Claude records only an answer you gave. It never picks one for you. More in questions and on your phone.

What Claude can do in Hamloy

All 13 tools: list and read items, add tasks, requests and questions, claim and tick, edit and move, record your answers, message the agent on a task, read the news, and make views of the board. Since Claude keeps one connection for your account, it reaches every project you're on, and you can ask it which ones there are or start a new one. See the tools.

What Claude in a chat doesn't get is what the hamloy program adds on a computer: hooks, git, the panel and the status line. Those are for Claude Code and Codex.

Connect in a minute

1. In Claude, open Customize › Connectors, choose + Add, then Add custom connector.
2. Name it Hamloy and give it https://hamloy.com/mcp as the MCP server URL.
3. Choose Continue, and sign in to Hamloy in the window that opens.

That one connection works on the web, the desktop and the phone. The full steps are in Claude setup and, for the desktop app, Claude Desktop setup.