Hamloy vs Claude Code's own to-do list
Claude Code's task tools keep one session's steps on your computer. Hamloy keeps a project's tasks for every session, tool and person. They fit together.
Checked against Claude Code's own site on 10 October 2026
The short answer
Claude Code's task tools keep a checklist of the steps of the work in front of it, on your computer, and come with Claude Code; on newer models they are off unless you turn them on. Hamloy keeps the project's open work for every session, tool and person, with claims, questions and requests for you, a board you can read anywhere, and done followed through git. They don't compete: Claude can keep its steps in its own list while it works on one Hamloy task.
| Hamloy | Claude Code | |
|---|---|---|
| What it holds | A project's tasks, requests for you and questions for you | A checklist of the steps of the work at hand |
| Where it's kept | On hamloy.com, hosted in the EU | In ~/.claude/tasks on your computer |
| How long it lasts | Until each item is done or deleted | Through compactions; a resumed session picks it up |
| Shared with | Every session of Claude Code, Codex, Cursor, VS Code, Claude and ChatGPT on the project, and the people you share it with | Claude Code sessions that set the same CLAUDE_CODE_TASK_LIST_ID, or one agent team |
| Two agents, one task | Claims, across tools and machines | In agent teams (experimental), teammates claim tasks with file locking |
| Seeing it | A board in the browser, a panel in the terminal, Claude Code's status line | Ctrl+T in the terminal, up to five tasks at a time; VS Code's Focus view |
| On by default | Yes, once the folder is connected | Only with older models; on newer ones you turn it on |
| Cost | Free for one project; Pro €8 a month per person | Built into Claude Code |
When to choose which
Claude Code's own list is enough when one Claude Code session does the work and you watch it: the list keeps the steps in order through a long task and a compaction, needs no account, and stays on your computer. With several Claude Code sessions on one machine, a shared list ID or an agent team goes further still.
Choose Hamloy when the work outlives a session or crosses a boundary that list doesn't: Codex or Cursor working beside Claude Code, a second computer, another person, a decision you'll answer from your phone, or "done" that has to mean merged.
Use both, which is what most people connected to Hamloy end up doing without trying: Hamloy holds the project's tasks, Claude's list holds the steps of the one it's on.
What Claude Code's list is
From Anthropic's docs, read on 10 October 2026:
- Four tools: TaskCreate, TaskGet, TaskList and TaskUpdate. TaskUpdate changes a task's status, dependencies or details, or deletes it (tools reference).
- Not on every model. The tools are offered by default on Claude 3.x,
Opus 4 to 4.7, Sonnet 4 to 4.6 and Haiku 4.5. The docs say that on newer
models "Claude keeps track of multi-step work without a written checklist";
CLAUDE_CODE_ENABLE_TODO_TOOLS=1turns the tools back on. The older TodoWrite checklist comes back withCLAUDE_CODE_ENABLE_TASKS=0. - Kept on disk in
~/.claude/tasks/, one folder per list (the .claude folder). Tasks persist across compactions, and Ctrl+T shows the list, up to five tasks at a time (interactive mode). - Shared between sessions when each sets the same
CLAUDE_CODE_TASK_LIST_ID, a named folder in~/.claude/tasks/(environment variables). - Agent teams, experimental and off by default: a lead makes tasks, teammates claim the next unblocked one, and "task claiming uses file locking to prevent race conditions". The team's list stays on your computer, one team per session (agent teams). The same page points to cross-session messaging, for passing findings between Claude Code sessions you run yourself.
Anthropic's docs don't say whether other tools, such as Codex or Cursor, can read this list.
What Hamloy adds
Hamloy is outside any one tool. Claude Code and Codex in two worktrees, Cursor in a third, and Claude on your phone all call the same lists. A claim holds across all of them (claims):
⏺ hamloy - Claim an item (MCP)(key: "B-k3f9")
⎿ B-k3f9@4 claimed by this session
⏺ I'm on the streak counter. Counting days in a row per habit, with tests for skipped and missed days.
• Called hamloy.tick({"key":"B-r5nd@3"})
└ [x] B-r5nd@4 Dark mode for the habit list
• Dark mode is done and ticked.
› Take the streak counter on the home screen.
1 agent at work · 3 tasks ready
Being worked on
Ready, nobody on it most important first
• Called hamloy.claim({"key":"B-k3f9"})
└ B-k3f9 is claimed by claude since 4m ago; pick another item (B-vaxb is free), message them with the message tool (to: B-k3f9), or pass force to take it over
• Claude has the streak counter. Hamloy says B-vaxb is free, so I'll take the haptic tap.
• Called hamloy.claim({"key":"B-vaxb"})
└ B-vaxb@2 claimed by this session
Things that are yours stay on the lists until you deal with them. A question comes with options and the one the agent recommends; a request comes with exact steps and what it frees. You answer on the board, and the next session, of any tool, carries on:
⏺ hamloy - Add an item (MCP)(kind: "request", title: "Make an APNs key for push reminders", body: […], needs: ["R-h2qe"])
⎿ R-p4nk@1 Make an APNs key for push reminders
⏺ Push reminders need an APNs key, and only you can make one. The steps are on your list. I'm going on with the settings screen meanwhile.
- In developer.apple.com › Certificates, IDs & Profiles › Keys, press +
- Name it Pebble push and tick Apple Push Notifications service
- Download the .p8 file and put its path in .env as
APNS_KEY_PATH, its id asAPNS_KEY_ID
Doing this lets Push reminders at the time each habit is due go ahead.
Hamloy also follows the work through git: a task ticked on a branch says so until the branch is merged and pushed (git). In Claude Code you can see what waits for you in the status line (Claude Code).
How they work together
Connect the folder with the hamloy program and nothing about Claude Code's
own list changes. A session starts with the project's lists and its next task,
claims a Hamloy task, and, if its model keeps a task list, breaks that task into
steps there. When the steps are done it ticks the Hamloy task, and the others
hear it:
⏺ hamloy - Claim an item (MCP)(key: "B-k3f9")
⎿ B-k3f9@4 "Streak counter on the home screen" claimed by this session
⏺ I'm on the streak counter. Counting days in a row per habit, with tests for
skipped and missed days.
The rule of thumb: if another agent or you will ever ask "is anyone on this?", it belongs in Hamloy. If it is step 3 of 7 inside one task, Claude's own list is the right place.
A plan already written down, in a PLAN.md, a TODO file or what an earlier session left in notes, comes in the same way: ask the agent to bring it in (import), and it lands in one call.
Questions
Does Hamloy turn off Claude Code's own list?
No. It changes nothing about Claude Code's tools. Claude can keep its steps in its own list while it works on a Hamloy task.
Can Codex see Claude Code's task list?
Anthropic's docs don't say so, and the list is kept in Claude Code's own folder. Hamloy's lists are the same for Claude Code, Codex, Cursor and the rest.
Isn't an agent team enough for parallel work?
Inside one Claude Code session, often yes. Agent teams are experimental, one team per session and on one computer. Hamloy is for sessions you start yourself, in different tools, on different machines, and for the decisions that come back to you.
Where does the status line come from?
The hamloy program can add one to Claude Code, if you want it. It shows the task the session is on and what waits for you. Codex has no status line.