Agentdockyard

AgentDockyard

**A local task hub for AI coding agents.**

AgentDockyard helps you keep track of what your AI agents are doing **without reopening every session**.

It works when you have **many different agents** working together, but also when you have **just one agent spread across many sessions**. You can immediately see what is **in progress**, **blocked**, **waiting**, **done**, and **who claimed what**.

Install the app, copy the prompt snippet shown in the built-in guide, paste it into your agent memory, and you're done. After that, your agents can create, update, claim, delegate, and close tasks on their own.

Local-first. No cloud dashboard. No account. No telemetry. By [Steeve Cordier](https://sitecrea.fr/) — MIT License.

---

Why this exists

When you run several AI sessions in parallel, the real problem is not generating code. It is **remembering the state of the work**.

Questions pile up fast:

• Which session is still working on something?
• Which task is blocked and needs input?
• Which task is finished already?
• Which agent already picked up