Agentweaver

Appearance

Agent definition ​

Every project you create in Agentweaver comes with a ready-to-use GitHub Copilot agent that knows how to drive the platform. You don't write it, register it, or keep it up to date — Agentweaver generates it from the live MCP tool set and drops it into your project automatically.

This page covers the experience: where the file shows up, how to use it with GitHub Copilot, and why it won't fight your edits. For the generation mechanics see the deep dive; for the exact contract see the reference.

What you get when you create a project ​

When you create a project — blank or cloned from GitHub — Agentweaver writes a single file into it:

.github/agents/agentweaver.agent.md

That file is the Agentweaver Driver agent. It contains a mental model of the platform (projects, blueprints, runs, the Coordinator, backlog, memory), operating principles, a Tool map of every agentweaver-* MCP tool grouped by category, and step-by-step playbooks (submit and supervise a run, stand up a project and team, work the backlog, curate memory and decisions).

It is written once, only if the file is not already there. A blank project always gets it; a cloned repo gets it unless that repo already ships its own .github/agents/agentweaver.agent.md.

Using it with GitHub Copilot ​

GitHub Copilot automatically discovers agent files under .github/agents/. After your project is created:

  1. Open the project's working directory in an editor with GitHub Copilot.
  2. In Copilot's agent picker, choose Agentweaver Driver (its description tells Copilot to use it when you mention Agentweaver, "spin up a team", "run a workflow", a project / blueprint / run / coordinator, or any agentweaver-* tool).
  3. Ask it to do platform work in plain language — e.g. "spin up a software-development project and run the software-delivery workflow." The agent translates that into the right sequence of agentweaver-* MCP tool calls, supervises the run, and reports back.

The agent file is just markdown — open it any time to read exactly which tools and playbooks the agent knows.

It won't overwrite your edits ​

The file is yours once it lands. If you customize it — tighten the playbooks, add house rules, change the description — Agentweaver will not clobber your version. Materialization only writes when the file is absent, so your edits are safe across anything that re-touches the project.

If you ever want to reset to the shipped version, delete your copy and recreate the project (or copy the current definition from the repo's own .github/agents/agentweaver.agent.md).

How the shipped definition stays correct ​

You never have to worry about the agent's Tool map going stale against the real tools:

  • The Tool map is generated from the MCP server source, so it always lists the tools that actually exist.
  • The copy embedded in the app and the copy in the repo are kept byte-identical, and CI fails the build if either drifts.
  • New projects always get the current definition.

So the agent a fresh project ships with is always in step with the platform it drives. For the full list of tools it can call, see the MCP tool index.

What to expect ​

  • It's automatic. No registration step — create a project and the agent is there.
  • It's per-project. Each project gets its own copy under that project's .github/agents/, so it travels with the repo and is committed alongside your code.
  • It's non-destructive. Existing agent files are never overwritten.
  • It's not a web-UI feature. There's no button or dialog to manage it — it's a file that appears in your project. Manage it like any other file in your repo.