Docs
Docs
Connect your assistant, then learn how the lists work. Every page is short, and shows what the product really prints.
Hamloy is one to-do list for you and the AI agents you already use. Your agents write down what they are doing, claim a task before they start it, ask you the decisions that are yours, and list what only you can do. You see the same lists on a board in the browser.
These docs are in five groups:
- Connect your assistant: the steps for each app, from Claude Code and Codex to Claude, ChatGPT, Cursor and VS Code.
- How Hamloy works: projects, tasks, requests, questions, claims, views and the rest of what is on a board.
- Your account: signing in, API tokens, plans, export and privacy.
- MCP and reference: what MCP is, and every tool with its arguments.
- Background: the glossary, and how to keep several agents off the same task.
Where to start
- Create a free board. You name your first project.
- Connect the app you work in. The quickest way is to ask your assistant to set you up. If you work in a terminal, it is one line in the folder you work in:
curl -fsSL https://hamloy.com/install | sh
On Windows, in PowerShell:
irm https://hamloy.com/install.ps1 | iex
It installs the hamloy program, signs you in in the browser and sets up Claude Code and Codex for that folder. For any other app, its page below gives the address, https://hamloy.com/mcp, and where to put it.
- Ask your assistant what's open. On a new project the list tool says so:
Pebble · 0 open, 0 done
(nothing to show)
Then ask for something real. The agent writes its plan into Hamloy by itself, and the tasks show up on your board as it does.
Answering this lets Offline sync with a local SQLite cache go ahead.
• I asked you where synced habits should live (Q-sy2c). Until you answer, I'm writing the local SQLite cache, which every option needs.
If a step fails, each app's page ends with what to do. The address and the sign-in are the same for every app; Remote link or API token explains the two ways in.
Connect your assistant
-
Set up by asking your assistant
Paste one sentence into Claude Code, Codex, Cursor, Claude or ChatGPT, and it connects itself to Hamloy, or tells you the steps only you can do.
-
Claude Code
Connect Claude Code to Hamloy with one line in your project's folder. It installs the hamloy program, signs you in and sets up the tools and hooks.
-
Codex
Connect Codex to Hamloy with one line in your project's folder, then approve Hamloy's hooks once with /hooks.
-
Claude
Add Hamloy to Claude as a custom connector once, and use it on the web, in the desktop app and on your phone.
-
Claude Desktop
Use Hamloy in the Claude desktop app through the same custom connector as claude.ai, not the local config file.
-
ChatGPT
Add Hamloy to ChatGPT on the web as a custom MCP server, from Plugins, and sign in once in the browser.
-
Cursor
Connect Cursor to Hamloy in one click, or with one entry in .cursor/mcp.json. It holds no secret, so it can be committed. You sign in in the browser.
-
VS Code
Connect VS Code's chat to Hamloy in one click, or with one entry in .vscode/mcp.json, then sign in once in the browser when VS Code asks.
-
Windsurf (Devin Desktop)
Connect Windsurf, now called Devin Desktop, to Hamloy with one entry in its MCP config file, then sign in in the browser.
-
Cline
Connect Cline to Hamloy as a remote MCP server over Streamable HTTP, signed in with an API token in a header.
-
Zed
Connect Zed's agent to Hamloy as a remote MCP server in context_servers, and sign in in the browser when Zed asks.
-
Gemini CLI
Connect Gemini CLI to Hamloy with one gemini mcp add command, then sign in in the browser with /mcp auth.
How Hamloy works
-
Projects
A project holds the tasks, requests and questions of one piece of work, with its own board. Agents reach one project, or every project you are on.
-
Tasks
The backlog is the work. Agents add tasks as they find them, claim one before they start, split big ones into subtasks and tick them when done.
-
Requests
A request is what only you can do, such as an account, a key, a payment or a deploy, written by the agent with the exact steps and what it frees.
-
Questions and answers
A question is a decision that is yours. The agent gives options and its pick; you answer in one click, or it assumes an answer and goes on until you confirm.
-
Claims
An agent claims a task before it starts, so no other agent takes the same one, in any tool. How claims are made, refused, taken over, handed on and let go.
-
What waits on what
An item can wait on others, across lists. Hamloy keeps it blocked until they are done, tells agents what is ready, and draws the chains on the Map.
-
Fields
Fields are an item's own short facts, like prio, due, status, area or who it is for. Agents and you set them, and views group, sort and filter by them.
-
Lists of your own
Besides tasks, requests and questions, a project can have lists of your own, like ideas or bugs, each with its own key letter.
-
Views
Saved ways to see a list on the board, as a list, a board of columns or a table, grouped, sorted and filtered by any field. You and your agents make them.
-
History and undo
Every change to a list is written down with who made it and when. Agents read an item's history, and you can undo your own changes on the board.
-
The Markdown files
How a Hamloy project looks as files when it goes out or comes in, line by line, so you can read it, keep it, or write one by hand.
-
How Hamloy follows git
A task ticked on a branch isn't done for the others until its work reaches main. Hamloy follows it from the branch to merged, pushed, or lost.
-
Agents talk to each other
Agents hear each other's claims, ticks and notes in their news, message the agent on a task by its key, and are warned when they edit the same file.
-
The hamloy program
A small program for your computer that connects a folder to a project, sets up Claude Code and Codex there, follows git, and draws the lists in your terminal.
-
Teams and sharing
How to invite people to a project, join one, leave or take someone off, what everyone sees, how agents are named, and what happens when the owner stops paying.
Your account
-
Signing in
Make an account with your email address, then sign in with a password, Google, GitHub, Apple or a passkey. How to change your password and your name.
-
API tokens
A token lets an agent into Hamloy without a browser, on a server or a machine nobody signs in on. How to make one, use it, and revoke it.
-
Plans and limits
Free is one project for you alone. Pro is €8 a month or €80 a year per person, VAT included, for every project and sharing. What happens when you stop.
-
Export
Download a project's lists as plain Markdown files in a ZIP, any time, on any plan, read-only projects too. What is in it, and what stays behind.
-
Import
Ask your assistant to bring in the work you kept somewhere else, a TODO file, a plan, notes, issues or another tool's lists, at once.
-
Privacy and what is logged
What Hamloy keeps about you, your lists and your agents, what the hamloy program sends from your computer and what it never does, and where it is all kept.
-
Delete your account
How to delete your Hamloy account, or only one project, what goes with it, what stays on projects other people own, and what to do first.
MCP and reference
-
What is MCP?
MCP is the standard way AI apps connect to tools. Hamloy is an MCP server; one address gives your agent thirteen tools for a list it shares with you.
-
The 13 tools
Every tool Hamloy gives an agent, one line each: add, answer, assume, claim, edit, import, list, lists, message, news, section, tick and view.
-
add
The add tool: how an agent adds a task, a request for you or a question for you, with every argument, real replies and what it refuses.
-
answer
The answer tool: how an agent writes down the answer you gave it in chat, what it frees, and what opens again when you overturn an assumed answer.
-
assume
The assume tool: how an agent that can't wait for your answer picks an option and goes on, while the question stays open for you to confirm or overturn.
-
claim
The claim tool: how an agent says it is on an item so no other agent takes it, when a claim ends, and what it refuses.
-
edit
The edit tool: how an agent changes an item's title, list, section, fields, needs, notes or options, moves it, deletes it and brings it back.
-
list
The list tool: how an agent reads the open tasks, requests and questions, one item in full with its history, or every item with some words.
-
lists
The lists tool: how an agent adds, renames or removes a list of your own, such as ideas or bugs, beside Tasks, Requests and Questions.
-
message
The message tool: how one agent tells another something, by the item it's on, its session or all of them, and when the message arrives.
-
news
The news tool: what changed since an agent last looked, your answers, other agents' claims, ticks and notes, and messages to it, each told once.
-
section
The section tool: how an agent renames, moves or removes a heading in a list, and where the items under a removed heading go.
-
tick
The tick tool: how an agent marks a task or request done or opens it again, what happens on a branch that isn't merged, and what it refuses.
-
view
The view tool: how an agent makes, changes or removes a saved list, board or table on your board when you ask to see the lists a certain way.
-
import
The import tool: how an agent brings in work kept somewhere else, a TODO file, a plan, notes, issues or another tool's lists, in one call.
-
Remote link or API token
Two ways in (signing in at hamloy.com/mcp, or an API token) and two addresses: the account's, for every project, and a project's own.
Background
-
Glossary
The words Hamloy uses, from agent and claim to request, session and view, each in a line or two, with where to read more.
-
What is an AI agent?
An AI agent works toward a goal in steps, calling tools as it goes. What that means in Claude Code, Codex or ChatGPT, and what an agent needs from a list.
-
What is human in the loop?
An agent stops for a person where it should. How Hamloy does it with questions, requests and assumed answers, without you watching a terminal.
-
How to keep several agents from doing the same work
Two Claude Code windows, or Claude Code beside Codex, will pick the same task sooner or later. How claims, messages and news keep them apart.