For humans and coding agents

Turn project context
into verified work.

Ask your coding agent to run a scoped epic. Docket keeps the tasks, project knowledge, and verified outcomes together in Git—ready for your review and the next session.

Local-first · Plain Markdown · Reviewed through Git

Pick up where you left off

The project, in context.

See what changed, what matters, and what’s ready next. The interface reads the same files your agent works with.

Current GitDocket Home: Home and Board lead the navigation; a project re-entry note summarizes the completed Harbor guide, and a compact board shows feedback ready next.
Current development UI · Synthetic Harbor project · View full size
Explore the current Board view
Current Harbor Board with one todo task, two done tasks, collapsed empty columns, and filtering controls.
Filter the work and collapse empty columns to keep the next step visible.

A repeatable loop

Finish the work. Keep the knowledge.

“Run the welcome-guide epic through completion.”

  1. 01

    Read & choose

    Read project context and select a scoped task with completed dependencies.

  2. 02

    Build & verify

    Make the change and check it against the task’s acceptance criteria.

  3. 03

    Close & carry forward

    Reconcile docs, record Git-linked evidence, and leave the next step ready.

Your agent follows the checked-in procedures, reviews the whole epic, and returns a completion receipt or a concrete blocker. You direct the work and review the result.

Inspect a scripted run with real CLI evidence

An open foundation

Your knowledge stays yours.

Built on Open Knowledge Format v0.1, introduced by Google Cloud. Linked Markdown holds the context; Docket adds tasks, dependencies, agent workflows, and evidence.

Read it in your editor. Review it in Git. Bring it to your next agent session.

Start with a small epic