Should you choose Task Master or ConvOps?
Choose Task Master when you need a backlog planned from a PRD. Choose ConvOps when the agents you already use must follow a process with sign-off, memory and a record.
choose Task Master if
- You start from a PRD and want it broken into tasks with dependencies and subtasks.
- You want "what is the next task?" answered inside Cursor, Windsurf, VS Code or Claude Code.
- You prefer a local tool: an npm package and a task file in your project.
- You want to hand a task list to Claude Code and step away while it loops through them.
choose ConvOps if
- Each task should follow the same steps every time, with sign-off where it matters.
- Work moves between AI clients and people, and must pick up where it stopped.
- Decisions and their reasons should be remembered for the next session.
- You need an audit trail and governance across teams, not only a backlog.
How do Task Master and ConvOps compare?
Side by side, as of October 2026. ConvOps is the operations layer for AI agents: an MCP server that holds the process your AI clients follow, with approval gates, a shared memory and an audit trail. Every Task Master cell links to the official source it comes from; ConvOps cells describe the product as it ships today.
| Capability | Task Master | ConvOps |
|---|---|---|
| What runs the steps | Your editor's AI works the tasks. Task Master's own AI commands (parse, expand, research) call a model you configure, or Claude Code or Codex CLI without an API key. [1] | Your AI client runs each step. ConvOps hands it one step at a time and runs no AI models itself. |
| Where the process lives | A task list in tasks.json in your project. Team mode syncs tasks to Hamster's cloud. [3] | A workflow in your ConvOps workspace, reached over MCP and changed by asking your AI client. |
| Human approval gates | Its docs describe task status, dependencies and subtasks, not approval gates. Decisions stay with you and your editor. [2] | Approval gates on any step, evaluated by the engine. The run waits at the gate, for minutes or days, until approval is passed. |
| Memory | Context lives on each task: description, details, test strategy and subtasks. The research command pulls fresh information with project context. [2] | A shared brain of memories and routes, recalled across sessions, AI clients and workspaces. |
| Audit trail | Changes are edits to the task file in your project. Team mode stores tasks in the cloud and syncs them in real time. [3] | An audit row for every change to tasks and workflow runs, keeping the agent's self-declared label apart from the server-verified human. |
| Attended or unattended | Attended in your editor. The loop command runs Claude Code through the next available tasks while you step away. [4] | Attended in your AI session, or unattended through executors and RRULE schedules, including one pod per run on your own Kubernetes. |
| Self-hosting | Runs locally as an npm package and MCP server (npx task-master-ai). [1] | Hosted at mcp.convops.app, or self-hosted on your Kubernetes with one Helm chart on Enterprise. |
| Pricing model | Open source under MIT with Commons Clause. You pay for model API keys, or use Claude Code or Codex CLI. Hamster, used for team mode, has its own pricing. [6] | Free $0, Pro $79/mo ($66/mo billed annually), Business $299/mo ($249/mo annually), Enterprise from $2,500/mo. Every feature on every plan; you pay for scale. Your AI client is billed separately. |
| Client support | MCP setup documented for Cursor, Windsurf, VS Code, Amazon Q Developer CLI and Claude Code. [1] | Any MCP client. Verified: Claude Code, Claude (web, desktop and Cowork), OpenAI Codex, ChatGPT, Cursor and OpenCode. |
What is Claude Task Master?
Task Master is an AI-powered task management system for AI-driven development. It parses a PRD into structured tasks with dependencies and subtasks, then tells your AI editor what to work on next.
It runs as an MCP server or a CLI. You write a PRD at .taskmaster/docs/prd.txt, ask your editor to parse it, and Task Master generates tasks; from there you ask for the next task, expand one into subtasks, or run a complexity report (Task Master README). Each task carries an id, title, description, status, dependencies, priority, details, a test strategy and subtasks (task structure).
It works in solo mode, with tasks stored in your project, or in team mode, which connects to Hamster for cloud storage and real-time sync (team docs). The license is MIT with Commons Clause: you can use and modify it, but not resell Task Master itself, offer it as a hosted service or build competing products based on it.
Check the release pace before you commit. As of October 4, 2026, the latest task-master-ai release on GitHub is 0.43.1, published March 31, 2026 (releases).
What is the main difference between Task Master and ConvOps?
Task Master decides which task comes next. ConvOps decides how each task is done: the steps, the gates and what gets recorded.
A Task Master task is a description plus a test strategy. How it gets done is up to the prompt and the model on the day. That is fine for a solo build and a weak spot for a team: two people asking for the same task can get two different processes.
In ConvOps, a task gets a workflow when it is created. The AI receives one step at a time, approval gates hold the run where a person must decide, and every change is written down. The task list still matters; ConvOps has tasks, child tasks and work pools too. The difference is that a process travels with every task. See agentic workflows explained.
Where is Task Master stronger?
In turning a document into a plan. ConvOps has no PRD parser.
PRD to task graph
parse_prd builds a dependency-ordered backlog from one document in one call.
Complexity and expansion
analyze_project_complexity scores tasks, and expand_task breaks the hard ones into subtasks.
Research built in
The research command queries fresh information with your project as context, with a separate research model.
Local and light
An npm package with a file in your repo. Tool loading modes cut its MCP footprint from 36 tools to 7.
Hands-off loop
The loop command runs Claude Code through the backlog task by task.
Where is ConvOps different?
It governs the work inside each task, for the whole team, in any MCP client.
A process per task
Steps arrive one at a time, so the agent cannot skip to the end or invent its own order.
Approval gates
The engine evaluates each gate on advance. The run waits until approval is passed.
Memory that compounds
Decisions and where things live are saved as the team works, then recalled by the next session in any client.
Record and roles
An audit row per change with the verified human, three roles and workspaces per team or customer.
Not only coding
Content, sales and operations tasks run the same way, including in ChatGPT and Claude desktop.
Can you use Task Master and ConvOps together?
Yes. Both are MCP servers, so one Claude Code session can load both: Task Master plans the backlog, ConvOps runs each task through a workflow.
Ask Task Master for the next task, create it as a ConvOps task, and let its workflow drive: plan, approval, build, review. When the workflow completes, mark the task done in Task Master. We have not built an integration between the two; this is two MCP servers in one session, each doing its own job.
one ConvOps workflow, with Task Master inside one step
How do you add a workflow to every task?
Connect ConvOps next to Task Master, describe your task workflow once, and start tasks from it.
- 1
Connect ConvOps
In Claude Code, one line. Task Master can stay connected as it is. Using Cursor instead? Follow the Cursor setup guide.
claude mcp add --transport http convops https://mcp.convops.app/ - 2
Keep Task Master for planning
The README's Claude Code install stays the same.
claude mcp add taskmaster-ai -- npx -y task-master-ai - 3
Describe the task workflow
Ask Claude: "Create a ConvOps workflow for feature tasks: plan, approval, build, review, done." Add a gate where you want to sign off.
- 4
Run tasks through it
For each next task, create a ConvOps task with that workflow. The AI walks it one step at a time and the record builds itself.
What else do people ask about Claude Task Master and ConvOps?
Short answers to the questions that come up most, with the same sources as the table above.
Is Claude Task Master only for Claude?
No. Its README describes it as a task management system for AI-driven development that works with any AI chat, and documents MCP setup for Cursor, Windsurf, VS Code, Amazon Q Developer CLI and Claude Code. Its AI commands can use several model providers.
Does Task Master have approval gates?
Its documentation describes task status, dependencies and subtasks, not approval gates. ConvOps adds gates the engine evaluates before a run can move past a step.
Can ConvOps turn a PRD into tasks?
ConvOps has no PRD parser. Your AI client can read the PRD and create tasks and child tasks in ConvOps over MCP, but the dedicated parsing, complexity scoring and expansion are Task Master's strengths.
What does the Commons Clause in Task Master's license mean?
You can use, modify and distribute Task Master and build products with it, but you cannot sell Task Master itself, offer it as a hosted service or create competing products based on it. ConvOps is a commercial service with a free plan.
Can I run Task Master and ConvOps in the same Claude Code session?
Yes. Claude Code can load several MCP servers at once. Use Task Master to decide what comes next and ConvOps to govern how each task is done.
Where do these facts come from?
From the official documentation, repository and pricing pages for Claude Task Master, checked in October 2026. Anything we could not verify there is left out.
- [1]Task Master README on GitHub
- [2]Task Master docs: Task structure
- [3]Task Master docs: Team collaboration
- [4]Task Master docs: Loop command
- [5]Task Master releases on GitHub
- [6]Hamster pricing
Spotted something out of date? Tell us and we will correct the page and its date.