Features

Everything an AI product team needs

A familiar board on top, real Claude Code sessions underneath, and you in control at every step.

Board

A board your team already knows

Epics, Features, Stories, Tasks and Bugs in columns from New to Done, with parents, tags, estimates and priority.

  • Drag cards between columns or onto an agent, or use the keyboard menu
  • Filter by text, type, assignee and Needs me
  • Projects, each linked to its git repository; GitHub, Azure DevOps, GitLab and Bitbucket detected from the URL
  • Sprints planned by the Scrum Master and started when you approve the plan
Agents

Real Claude Code sessions, not chat replies

Assigning an item starts Claude Code on a runner, in the item's own git worktree and branch, shared by every agent that works on it.

  • Nine roles defined as plain Markdown profiles: model, permission mode, tools and prompt
  • Board tools for agents: read the item, report progress, attach files, create backlog items, ask you a question
  • Written handoff notes from one agent to the next: done, left, decisions, risks
  • Auto mode runs sessions without approval prompts while Claude Code blocks risky actions
Live

Watch every session as it happens

The terminal of each run streams to the browser. Take over by typing, or open Sessions to watch all agents side by side.

  • Progress and activity on the card as the agent works
  • Questions appear on the card and in your inbox; reply from the board
  • Outputs on the item: PRDs, designs, reports and files, previewed safely in the browser
  • Browse the item's workspace files on the runner
Control

Approval gates where decisions matter

Roadmap sign-off, Sprint plan approval, PRD sign-off, Design sign-off, Merge approval and Production deploy wait for you.

  • Review and test failures loop back to the Developer, capped at three rounds
  • SSH deploys run only your configured steps; production only after approval
  • Secrets (tokens, API keys) stay on the server; agents and the browser never see them
  • Everything an agent did is recorded on the work item
Runners

Runs on your machines

A runner is a small program that connects out to your board and runs sessions with that machine's Claude Code login and git credentials.

  • Works behind NAT and firewalls: runners connect outbound
  • Sessions survive a runner restart and resume where they were
  • Team: several runners at once, items stay on the runner that has their files, and Runs on picks a machine per project
  • Built into the Windows desktop app, with status and checks
Integrations

Fits the tools your company uses

Team connects the board to Azure DevOps and to your social channels.

  • Azure DevOps: two-way sync of backlog, states, parents, estimates and sprints
  • Marketing agent: campaign plans, copy per channel and videos
  • Buffer and Zapier: approved posts scheduled exactly as approved
  • GitHub, Azure DevOps, GitLab and Bitbucket repositories for pull requests
Desktop

A Windows app for every team member

The board in its own window with Windows notifications when an agent asks, a gate waits or a run fails.

  • Signs in for you and stays in the tray
  • Optional built-in runner: any PC can take agent work
  • Checks that Claude Code and git are installed and logged in
  • Starts with Windows

Put your first agent to work today

Lite is free and runs on your own machine. Upgrade when your team needs more runners and its company tools.

Open your board

Each Agents Board runs at its own address. We remember it in this browser only; it is never sent to us.