mblua/AgentsCommander

Stop letting your experience disappear between sessions. Build it into your own agent ecosystem. For developers working solo or in Teams: keep roles, skills, and recorded experience in your repository, refine them from real work, and bring the setup to your next task. Your coding agents run in real terminals; you decide how they work together.
Explore the ecosystem
 · 
Get started
 · 
Your first team
## Get started
Bring your own coding-agent CLIs, accounts, and model access. Choose how you want to install AgentsCommander.
### 1. Install with npm
Requires Node.js 18 or newer and npm. Run:
bash
npm install -g @mblua/agentscommander@latest
Then launch:
bash
agentscommander
See the npm guide for supported platforms, verification, and preserving your configuration when upgrading.
### 2. Install with winget on Windows
powershell
winget install --id mblua.AgentsCommander --exact --source winget
### 3. Ask your coding agent
Paste this into the coding agent you already use:
text
Install the stable version of AgentsCommander with npm install -g @mblua/agentscommander@latest.
Verify that it works, tell me how to launch it, and explain the basic concepts used by AgentsCommander: Agents-Matrix, Teams, and Rooms.
Prefer a release download? See GitHub Releases and the installation guide for the right asset and verification steps. macOS uses the npm route.
Platform support: Windows is the primary supported platform. Linux has medium test coverage; macOS uses npm and has very low test coverage. Check the platform details before installing. Before updating an existing installation, follow the configuration backup guidance.
## Your experience belongs in your next task
A useful session leaves more than code: a clearer role, a tested skill, a recorded mistake, or a better way to delegate. Keep those lessons in files you can review and improve, then bring the setup to your next task.
AgentsCommander is an open source playground and meta-harness for coding agents. Your agents keep their own tools and run in real terminals. AgentsCommander adds the surrounding workspace for roles, skills, reusable Teams, Rooms, and coordination you can inspect.
Start with one agent configured your way, or build a Team with an orchestrator and the specialists you choose. You bring your CLIs, accounts, and model access.
## Create → Use → Refine → Reuse
1. Create your setup. Define agents, their roles and skills. Use one agent day to day, or group agents into a Team.
2. Use it on real work. Activate a Team as a Room, give its orchestrator a task, and follow the terminals and messages. Open another Room for another task.
3. Refine it from evidence. Review messages, recorded errors, plans, notes, tests, and diffs. You can do this yourself or prepare a team of agents for that review.
4. Reuse what works. Improve the roles, skills, and instructions you maintain, version the reusable configuration, and carry it into the next task.
You decide what to record and improve. This workflow preserves explicit knowledge in files; it does not automatically train models or guarantee better results.
## Start with one agent. Grow your own ecosystem.