Hypertask Alternative: Hypertask vs Weft, Compared Fairly

Pick Hypertask if you want a keyboard-fast board with AI built in, a CLI for scripts and CI, and a REST API for your own workers. Pick Weft if your agents live in ChatGPT, Claude or Cursor and should claim tasks, ask you questions and hand back evidence through MCP itself, with native phone apps and a $7 or $60-a-year seat. Both are agent-native boards. They put the agent plumbing in different places.

We make Weft. Facts about other products come from their official pages, checked on 27 October 2026.

What do Hypertask and Weft both do?

More than most task tools. Both run a hosted MCP server over Streamable HTTP, so an AI client can read and change the board without anything installed locally. Both treat agents as real participants: tasks show who did what, whether a person or an AI client. Both have a free plan that includes MCP access. And both are built for people who already work with AI agents every day, not for adding a chatbot to a legacy tracker.

So this is not a comparison between a modern tool and an old one. It is a comparison between two products that answer the same question differently. If you like Hypertask, most of what you like will still be true after reading this.

How do Hypertask and Weft compare, row by row?

Every Hypertask cell links to the Hypertask page it comes from. Where a page says nothing, we say so.

Two rows need a word of context. On sign-in, Hypertask does support a browser approval flow; its live server publishes standard OAuth metadata with dynamic client registration and PKCE. The difference is that its documented setup is a copied API key that expires after 30 days, and a browser-approved connection lasts 90 days without a refresh token. On leases and evidence, Hypertask has both. It deliberately keeps them in its REST API for autonomous workers, rather than in the tools a chat assistant sees.

When is Hypertask the better choice?

In several real situations. We would point you to Hypertask if any of these is what you care about most:

• Keyboard-first speed. Hypertask is designed for people who move, comment and close dozens of tasks an hour without touching the mouse (source). That is a headline of their product. We do not make that claim for Weft.
• AI built into the board. Describe a task in plain English and AI fills in the title, description and tags. Smart labels keep classifying new and edited tasks against a rule you write. Natural-language filters such as “top-priority bugs due this week” work in the board filter (source). Weft has no built-in AI: it relies on the assistant you connect.
• A CLI for scripts and CI. The npm package @hypertask/hypertask_cli lists, moves and updates tasks from a terminal or a pipeline, with the token from an environment variable (source). Weft has no CLI.
• A REST API. Hypertask’s surfaces are wrappers over the same REST routes, and agent infrastructure such as leases and evidence is exposed there for autonomous workers (source). Weft has no public REST API and no webhooks.
• Agents as service accounts. In Hypertask an agent can be its own account with scoped API keys and a bot badge, and its writes are attributed to the agent, not its owner (source). In Weft each OAuth connection and each access token is a separate actor, and OAuth connections can be limited to full, write or read, but access tokens carry no scopes.
• Time tracking. Timers, a Time page and time entries you can edit over MCP and CLI (source). Weft records estimates and cycle time, not tracked time.
• More capability per tool. Hypertask groups its MCP surface into 21 action-based tools (source), which keeps the tool list short while covering a lot. Weft exposes 32 separate tools.

If you want managed AI without thinking about keys, Hypertask Pro runs on Hypertask’s models; if you prefer your own keys, BYOK costs less (pricing).

When is Weft the better choice?

When your agents already run in a general AI client and the board is where they coordinate with you. Specifically:

• You work in ChatGPT. Weft is a published app in ChatGPT’s directory and works on every ChatGPT plan, including Free. There is no custom GPT or OpenAPI schema to set up (ChatGPT setup).
• You do not want to rotate keys. You sign in with OAuth in the browser, and because Weft issues refresh tokens, the client renews the sign-in itself. You can disconnect any client in Settings, or limit it to write or read.
• You want the claim loop inside MCP. start_task gives a 20-minute lease, so two agents cannot take the same card. request_input asks you a question with options and keeps the claim. complete_task records the outcome, artifacts, what was not done and what is unknown. A chat assistant sees and uses all of this as ordinary tools, with no separate worker to build (how the loop works).
• You want a limit on your own review queue. When 5 items are waiting on you (10 on a team), agents are refused new work until you catch up. Most tasks go straight to Done; the ones that need review stop in your Inbox for Accept, Rework, Later or Drop.
• You review on your phone. Weft has native iOS and Android apps from the App Store and Google Play. A question from an agent arrives as a push.
• Your team does not all read English. Weft is available in 27 languages.
• Price per seat matters. Pro is $7 a month or $60 a year per seat, with no AI credits on top, because the AI work runs in the client you already pay for (pricing).
• You are moving from another tracker. Weft imports CSV from Linear, Jira and Notion, and Trello JSON, on every plan.

What Weft does not do: no CLI, no REST API or webhooks, no built-in AI, no time tracking, no GitHub or Slack integration, and no desktop app. If any of those is a requirement, Hypertask is the better fit today.

Can the same agent use both?

Yes, technically. Both are MCP servers, so Claude Code, Cursor or another client can have both connected at once. Weft’s setup pages cover 33 clients, including Claude Code, Cursor and Codex. In practice, one board as the source of truth works better than two, because agents decide what to pick up by reading the board.

How do I switch from Hypertask to Weft?

There is no direct Hypertask importer in Weft. Here are two approaches that work with what each product offers:

1. Sign up for Weft and connect it to the AI client you already use with Hypertask, so both servers are available in the same client.
2. Ask the assistant to read your open Hypertask tasks through Hypertask’s MCP server and list them back to you. Check that list before anything is written.
3. Ask it to create those tasks in Weft with bulk_create_tasks. That tool writes many tasks in one all-or-nothing transaction and counts as a single call against the Free quota.
4. Open the Weft board, check titles and columns, and keep Hypertask read-only until you are sure nothing is missing.

This is an approach, not a built-in migration: results depend on your client and how many tasks you have, and fields such as time entries or smart labels have no equivalent in Weft. If you would rather not involve an agent, recreate only the open tasks by hand. Finished work can stay in Hypertask as an archive. For other tools, see the MCP task manager comparison.

Is Weft cheaper than Hypertask?

Per seat, yes: Weft Pro is $60 a year or $7 a month, and Hypertask BYOK is $8 a month billed yearly or $10 monthly (pricing). But Hypertask BYOK includes built-in AI features on your own keys, and Weft has no built-in AI. Hypertask’s free plan has no task cap; Weft Free is 50 active tasks and 200 AI tool calls a month.

Hypertask has leases and evidence too. What is actually different?

Where they live. Hypertask keeps leases, evidence and escalation in its REST API by design (source), which suits autonomous workers you write yourself. Weft exposes them as MCP tools (start_task, report_progress, request_input, complete_task), so an assistant in ChatGPT or Claude uses them without extra code.

Does Hypertask support OAuth?

Yes, in practice. A browser-approved connection gets a 90-day access token without a refresh token (source). Its main documented setup is a copied API key that expires after 30 days. Weft uses OAuth with refresh tokens, so the client renews the sign-in itself.

Is Weft in Claude’s connector directory?

No. On claude.ai and Claude Desktop, Weft is added as a custom connector, the same way as Hypertask. In Claude Code it is one command (setup).

Does Weft have a CLI or API I can call from CI?

No. Weft has no CLI, no public REST API and no webhooks. If your workflow runs from scripts or pipelines rather than from an AI client, Hypertask fits better.

Can I try Weft without leaving Hypertask?

Yes. Weft’s free plan needs no card. Connect it alongside Hypertask in the same client and try it on one project first (docs).