Hamloy vs Task Master
Task Master turns a PRD into tasks with AI models and keeps them in your repo. Hamloy is a hosted list your agents and you share. When to use which.
Checked against Task Master's own site on 10 October 2026
The short answer
Task Master is the better choice if you want a plan generated for you: it reads a product spec, writes tasks with dependencies, breaks them into subtasks with an AI model it calls, and keeps them in tasks.json in your repo, open source and free apart from the model. Hamloy calls no model of its own: the agents you already use write the tasks, claim them, ask you what's yours, and you see it on a live board, across tools and machines. One writes the plan; the other is where several agents and you work from it.
| Hamloy | Task Master | |
|---|---|---|
| In its own words | See what every AI agent is doing, and what's waiting on you | A task management system for AI-driven development, designed to work seamlessly with any AI chat |
| Where the tasks live | On hamloy.com, hosted in the EU | .taskmaster/tasks/tasks.json in your repo |
| Making the plan | Your agent writes it with add: tasks, subtasks, what waits on what | parse-prd turns a spec into tasks; expand and analyze-complexity break them down |
| AI models | None of its own; your agents do the thinking | Calls a model you choose: an API key, or Claude Code or Codex CLI signed in |
| How agents reach it | 12 MCP tools; the hamloy program for Claude Code and Codex | An MCP server (36 tools with all loaded) and the task-master command line |
| Two agents on one task | A claim: the second is refused and told who has it | No claim described in its docs |
| What's for you | Questions with options and a recommendation; requests with steps | Not described; tasks can be set to review or deferred |
| Where you see it | A live board in the browser, a panel in the terminal, Claude Code's status line | The command line and your editor's chat |
| Source and price | Closed and hosted. Free for one project; Pro €8 a month per person | MIT with Commons Clause; free, you pay for the model |
When to choose which
Choose Task Master when:
- you start from a written spec and want it broken into tasks, subtasks and dependencies for you, with a complexity score for each;
- you want the plan in a file in your repo, next to the code, readable by any agent;
- you want it open source and local, paying only for the model;
- one agent works through the plan, in Cursor, Windsurf, VS Code or Claude Code.
Choose Hamloy when:
- several agents work at once, Claude Code and Codex in two worktrees, and must not take the same task;
- your agents should ask you the decisions that are yours, with options, and list what only you can do, and you'd rather answer on a board than in a terminal;
- the list should be the same on every branch and machine, with nothing to merge;
- you want to see done as merged and pushed, not only ticked.
What Task Master does
From Task Master's own README and docs, read on 10 October 2026 (github.com/eyaltoledano/claude-task-master, docs.task-master.dev):
- From a spec to tasks. You write a PRD, and
task-master parse-prdturns it into tasks.analyze-complexityscores them,expandbreaks them into subtasks, andtask-master nextshows the next task to work on. Tasks have a status, dependencies, a priority, details and a test strategy. - Kept in the repo, in
.taskmaster/tasks/tasks.json. Tags keep separate sets of tasks, for example one per branch. - Models: it needs at least one API key (Anthropic, OpenAI, Gemini, Perplexity for research, and others), unless you use Claude Code or Codex CLI signed in, which need none.
- MCP: added with
npx -y task-master-ai, with setups for Cursor, Windsurf, VS Code and Claude Code. The README says all 36 tools load by default; smaller sets of 15 or 7 tools save context. - Running in parallel: clusters group tasks that can run at once and start a Claude Code session that spawns subagents for them (clusters).
- License: MIT with Commons Clause, which forbids selling it or offering it as a hosted service.
Task Master's site now leads to Hamster, its makers' hosted product, where local tasks can be exported for a team (Task Master on Hamster).
What Hamloy does
Hamloy doesn't write your plan with a model; your agent does, as tasks with subtasks and needs, and the board's Map shows what waits on what. What Hamloy adds is what happens when more than one agent, and you, work from it. Each agent claims a task before it starts:
⏺ 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
What only you can do comes as a request with the exact steps and what it frees; decisions come as questions with options and a recommendation:
⏺ 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.
The hamloy program follows each task through git: ticked on a branch, merged,
pushed (git).
Using them together
They can share a project, but not a list: two places that both say what's next
will disagree. One way that works is to let Task Master write the first plan
from your spec, then ask your agent to bring tasks.json into Hamloy, and work
from Hamloy from then on. It comes in one import, with the
subtasks and what waits on what: Task Master's numbers get Hamloy keys, and
what waited on task 3 still waits on it. Most people will pick one.
Questions
Does Hamloy write tasks from a spec?
Not by itself. Give your agent the spec and ask it to plan the work in Hamloy; it adds the tasks, subtasks and what waits on what with the add tool. The thinking is your agent's.
Does Hamloy need an API key for a model?
No. Hamloy calls no AI model. Your agents sign in to Hamloy in the browser the first time; that is all.
Can several agents work from Task Master's tasks.json at once?
Its changelog mentions a fix for several Claude Code windows writing the file at the same time, and clusters run tasks in parallel inside one Claude Code session. Its docs describe no claims between separate agents. Hamloy's claims hold across tools and machines.
Are my tasks in my repo?
In Task Master, yes, in .taskmaster/. In Hamloy they are on hamloy.com, and you can export them as Markdown files any time.