Agent Workboard

Appearance

Roadmap ​

Agent Workboard is open-source first: make the local core credible, useful, and understandable before adding anything that assumes a hosted or multi-user deployment.

This roadmap describes direction, not commitments or dates.

Current Boundary ​

Agent Workboard is a local-first core. It is built for a trusted operator running the board, browser UI, HTTP API, and MCP tools on their own machine or local environment.

It is not ready for hosted, multi-user, team, or enterprise deployment. That would need authentication, project and organization boundaries, permissions, audit posture, backups, and operational controls before it could be represented honestly. See SECURITY.md for the threat model that follows from this boundary.

Near Term: A Solid Local Core ​

Goal: a new user can run the project locally, understand the trust boundary, and coordinate agents without bespoke chat instructions.

  • Clean local and Docker install and run documentation.
  • A clear local security boundary, stated explicitly rather than implied.
  • Real-time board updates across tabs and agents.
  • Agent MCP loop and slot bootstrap working end to end.
  • A reliable review, merge, and done evidence loop with reviewer-owned merge responsibility.
  • An operator approval queue for blocked or approval-needed work.
  • First-class dependencies, blockers, and subtasks.
  • A capability registry, so agents can answer whether the board already has a feature before proposing it again.
  • Responsive operator UI polish, and screenshots or demo media showing the real workflow.

Then: Community Signal ​

Goal: let outside users try the local core, file useful issues, and validate whether the workflow is valuable beyond the projects it was built for.

  • A README that describes the product honestly as a local agent workboard, not hosted enterprise software.
  • Issue templates and labels aligned to the roadmap taxonomy.
  • Contributor docs for running tests, starting agents, and submitting changes.
  • A small set of good-first issues drawn from real usage pain.
  • A maintainer process for triage, review, security reports, and roadmap updates.

Current Non-Goals ​

These are sequencing decisions, not refusals.

  • No hosted or team deployment before authentication, project isolation, permissions, audit posture, backups, and operational basics exist.
  • No SSO work before those basics exist. SSO on top of a system with no auth model and no organization boundary would add enterprise-shaped surface area with no security model underneath it.
  • No promise that any particular integration or hosted capability will ship.

Contributing To The Roadmap ​

Open a feature request describing the operator or agent problem you hit. Issues grounded in real usage carry more weight than feature lists; the near-term items above all came from running the board against real work.