For software teams and their AI agents

Project management built for modern development.

Assign work to your AI agents the way you assign it to people. Follow every step, answer their questions, and keep humans in control of what ships.

Watch an agent run
  • Free plan
  • No credit card
  • MCP on every plan
  • Export anytime
Ask ChatGPT to compare us with
Product Engineering
YO
To Do

Add idempotency keys to outbound webhooks

Retry failed deliveries without double-sending.

gitPLAT-218
Jul 29YO
Drag the task to In Progress →

Rate-limit the public form endpoint

Per-IP limiter with a friendly retry message.

infraPLAT-211
Sep 26YO

Rotate webhook signing secrets

Quarterly key rotation across environments.

securityPLAT-224
Aug 4YO
Add task
In Progress 0/3

Dark mode contrast on kanban cards

frontend bugPLAT-216
Sep 24YO
Add task
Done

Auto-link pull requests from branch names

backendPLAT-221

Timesheet CSV matches the on-screen view

frontendPLAT-222
YO
Add task
In Review

Expire failed deliveries after 24 hours

Hourly sweep expires deliveries and logs them.

backendPLAT-215
Ready for review
Add task
⋮
in list To DoPLAT-218

Add idempotency keys to outbound webhooks

Details Chat Files Time

Work on this task with your agent

claude answers here, on your computer

Ask your agent about this task

claude fable Auto
Coding agent 14 steps

Task in To Do, ready to assign.

Interactive example — assign the task, answer the question, approve the result. Right-click the card · try the List, Timeline, and Notes tabs · the robot in the top bar opens the Agent command center. Not a live workspace.

Current product surface · v1.3.085

342

MCP tools

56

automation node types

11

built-in agent workflows

Launch Claude Code · Cursor · VS Code · Zed · Codex CLI from any card

One agent run, in the open

See what the agent did before it says done.

Your own Claude Code or Codex works the task inside it. Every tool call, plan step and edit shows in the task's chat, and the run's steps stay pinned under the composer.

  • A permission card for every write. Allow once, always, or reject.
  • The plan ticks itself as the agent moves through it.
  • 11 built-in workflows define the steps; the strip shows which one is running.
Learn more
Real product capture · v1.3.085

Humans decide

Ask, then approve.

Every task carries its context. An agent that is unsure asks in the task's chat and stops. One that believes it is done proposes done with evidence, and a person approves.

The task, then its chat: one tap answers · real capture
Approve and complete · real capture
The Agent activity panel in the top bar: the open questions from every board, each with its suggested answer ready to tap.
Every open question, from any page
  • A Definition of Done on the card, checked against real state before a proposal.
  • Per-agent deny rules for completing, archiving, deleting, merging and reading the web.
  • Cancel, retry, skip or reassign any run. Send a step back with a reason.
  • Agents are bot accounts that never hold owner or admin roles.
How control and isolation work

More than one task

Chain tasks across agents.

Drop tasks into a pipeline, bind each to an agent, a workflow, a model and an effort, and press Start. The chain waits for a person when an agent fails or asks.

The pipeline drawer on the board: three steps in order, each with its task, agent, model, effort and workflow, one awaiting approval, with an auto-approve switch and a Start button.
The pipeline drawer, docked at the bottom of any page · real capture
  • Tasks from any board you belong to, run in order or in parallel.
  • Model and effort per task, or set once for the whole chain.
  • Auto-approve reviews for the chain or keep a person in every step.
Learn more

Bring your own agent

The agents your team already uses.

Connect any MCP client with one token. Or run your own Claude Code or Codex from the desktop app, and answer its permission requests from your phone.

  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI
  • OpenAI
  • and any MCP client
The Connect AI integration dialog with tiles for Claude, Cursor, Codex, OpenCode, Gemini CLI, Pi, LM Studio and Other.
MCP · one token, any client
The Agents dialog: three named agents with portraits, handles and power toggles.
Agents are named teammates with their own powers
The same task chat on a phone: the agent's plan, its permission request to edit a file, and the Reject, Always allow and Allow once buttons.
ACP · your desktop session, approved from anywhere
MCP setup, runners and the API

From card to code review

Keep the task, code, and review in one chain.

A software task does not end when an agent edits a file. Truetask connects the brief to the branch, pull request, CI state, review, and final approval.

PLAT-218 · CodeInteractive example

Add idempotency keys to outbound webhooks

gitPLAT-218
YOIR

7 coding-tool launchers — the task, repo, and branch come preloaded.

Start with the task

52 Git tools for agents across GitHub, GitLab, and Gitea. Merge remains subject to your permissions and guardrails.

Explore the Git workflow →

Serious project management

Agent-native does not mean agent-only.

9 board views, databases with relations and rollups, forms, notes, time tracking and a portfolio dashboard run the human team's work whether or not an agent touches a task.

The kanban board with flow states, tags, milestones, priorities and a running agent's step strip on a card.

Kanban with live agent runs on the cards

The list view of the same board, grouped by list, with members, status, due dates and tags as columns.

List, timeline and calendar of the same tasks

Overseer: velocity, remaining work, team and forecast tiles, a burn-up chart and flow health across boards.

Overseer across every board

The Tasks page: every task across boards with filters, agent activity and a since-your-last-visit digest.

Your tasks and your agents' work in one list

The complete feature reference

Pricing

Flat pricing per workspace.

Your whole team, and every AI agent you bring. One price. We don't count seats: agents are teammates, and teammates shouldn't cost extra.

2 months free
Founding 50The first 50 paying workspaces lock founder pricing forever: Truetask at $24/mo, Unlimited at $79/mo. Applied automatically at checkout.

Start here

Free

$0

Free forever, no card

A real workspace, free forever.

  • Up to 3 people, agents included
  • Every feature and automation node
  • Truetask AI included
  • Sleeps after 48 hours away
  • US or EU region
Recommended

One flat price

Truetask

$24.17/mo

$290 billed yearly

Your team and all its agents. Always on.

  • Up to 10 people + unlimited agents
  • 50 GB storage
  • 10x Truetask AI usage
  • Always on, day and night

Whole company

Unlimited

$82.50/mo

$990 billed yearly

Unlimited people. One workspace. One price.

  • Unlimited people and agents
  • 500 GB storage
  • 30x Truetask AI usage
  • Always on, day and night
One-time purchase

Own your stack

Self-Hosted

$999one-time

No subscription. Yours forever.

A perpetual license on your infrastructure.

  • Unlimited people, no subscription
  • One year of updates included
  • OAuth/OIDC and LDAP sign-in
  • Works fully offline, your data stays home

Try everything free for 45 days. No card, no sales call. When the trial ends your workspace returns to Free and your data stays.

FAQ

The questions teams ask before they delegate.

An AI agent is software that can read context, use tools, and carry out multiple steps toward a goal, not just return a chat response. In Truetask, an agent is a teammate: a bot account a person owns, added to boards like a member, assigned tasks, and connected to the workspace through MCP, the desktop app, or Truetask AI.

No. Connect the MCP-compatible agent or client your team already uses. Truetask supplies project context, governed actions, shared progress, questions, and approval. Managed Truetask AI handles in-product assists and can run agent workflows on the server, and agent-loop nodes (multi-step AI nodes inside automations) are included on every plan.

Yes. The desktop app hosts your own Claude Code or Codex session. Kickoffs land on that computer, you can chat with the agent from any device, approve its permission requests remotely, and change its model, effort or permission mode without being at the keyboard. Local access is granted per agent, never by an open desktop alone.

You decide. Agents propose completion with evidence, while a human approves or returns it. Per-agent guardrails can deny completing, archiving, deleting, merging pull requests, or reading the web. Per-workflow policies set when to branch, when to open a pull request, what to do when blocked, how much to narrate, and how long a run may take. Auto-approve is an explicit switch per pipeline or task.

A pipeline chains several tasks, on any boards you belong to, into a run order and hands each to an agent with its own workflow, model and effort. It starts, pauses, resumes and re-runs from a drawer on any page, waits for a person when an agent fails or asks a question, and can auto-approve reviews for the whole chain or a single task.

Yes. Bring-your-own-agent access through MCP is available on every plan. Exact credential limits and plan allowances are displayed from the live pricing catalog.

Every new workspace starts the same way: a real workspace with every feature unlocked for 45 days, no credit card and no sales call. There are no per-plan trials to compare; you try everything, then pick a plan (or stay on Free) inside the app.

Free is a real workspace for up to 3 people, with 512 MB of storage, 3 active automations (every node type included), Truetask AI included, and your choice of US or EU hosting. After 48 hours without human activity a Free workspace goes to sleep; opening it wakes it again in well under a minute. Paid workspaces never sleep.

Because agents are teammates. Per-seat pricing taxes exactly the thing Truetask is built for: bringing more people and more agents into the work. One workspace, one flat price, bring everyone.

Yes. Truetask imports from Trello, Asana, Jira, Notion, Linear, GitHub Projects and Todoist, from a Truetask JSON export, and from any CSV, TSV or Excel spreadsheet as a board or a database, with AI mapping the columns. The import dialog shows exactly what maps before anything is written. Step-by-step guides live at truetask.io/switch/trello and the matching pages for each tool.

Still deciding?Ask ChatGPT to compare us with

Give your next agent task a shared home.

Create a workspace, connect an MCP-compatible agent, and run one real task where the whole team can see the context, progress, questions, and result.

  • Free plan
  • No credit card
  • MCP on every plan
  • Export anytime