Features

Built around the plan your agent makes.

Pages to plan from, a run you can follow, questions that don’t stop the work, and a backlog that closes with review. Each has its own page below.

Plan

Documents for specs and plans.

Specs, research and notes live in Documents, next to the plans your agent records. It reads them before it plans; you see every version.

More on Documents
Dana writes a brief in Documents, specs/sign-in, and saves it. The coding agent reads it before planning, then records its plan next to it as plans/sign-in. Dana tightens the brief to a second version, and the agent writes that into the plan as its second version. Every save is kept, with who made it.

Run

Follow the run from any browser.

Each agent session is a run on Now: the task in hand, the checks, what it's waiting on you for. It says Quiet when the agent stops calling in.

More on Now
A run on the Now page. Each call the agent makes is a heartbeat: start_task, two checkpoints with checks at 38 of 42 and then 42 of 42, and finish_task. The run is on track. Then the calls stop. After 30 minutes without an update it says Quiet, with the time of the last one. After 4 hours it moves to Earlier. A new session starts, and a new run goes live on top.

Ask

Keep the work moving while you decide.

The agent asks with a default and carries on. Costly questions go first, blocking ones hold only their tasks, and your answer reaches it on its next call.

More on Decisions
Waiting on you, next to the agent's task list. Three questions arrive, each with a default. The costly one sorts to the top. A blocking one holds the task it affects, and the agent moves to another task. You answer it differently from the default: the task is unblocked and marked until the agent hears your answer on its next call, then it goes back. You keep the default on the costly one. One question stays open, its default in use.

Review

Close every task with a review.

Epics and tasks in priority order, an icebox for later, and Ready for Review before anything is done. Reject with a reason the agent reads.

More on Backlog and review
A backlog in priority order, with an icebox beside it and review and done to the right. Five tasks settle in, the top one started. Audit log export is promoted from the icebox to fourth. The first task finishes and waits in review; you accept it and it's done. The second waits in review; you reject it with a reason, and it goes back to the top of the backlog, started, with the reason on it.

Team

Your whole team, and every bot on it.

Workspaces belong to your organization, so everyone in it sees them. Admins invite by email and manage seats. Each bot gets its own named token, so its work shows up under its name.

  • Dana Admin
  • Sam Member
  • Coder bot token
  • Reviewer bot token

Team is $15 a seat each month, with 250 AI credits per seat. Pricing

For developers

MCP, REST and a skill file.

Beck’s MCP server runs in Claude Code, Cursor or any MCP host. The REST API takes the same token, and agents read the rules at /skill.md.

install
npx @becklabs/beck-mcp-server
# BECK_API_TOKEN=beck_…  BECK_API_URL=https://beck.bot

Connect an agentMCP reference

Start with one plan.

See plans

Free includes 50 AI credits. Recorded plans, runs and decisions use none.