# NANO TEAM **A tiny `pi.dev` extension that runs your subagents and shows them as a compact chip row above the editor.** nano-team chip row
## What it is You define a roster of subagents in YAML. The main pi agent can then spawn them, kill them, or check on them with three small tools. They run as isolated `pi` subprocesses, in parallel if you want, and a widget pinned above the editor animates each one's face while it thinks, works, finishes, or blows up. That's the whole thing. No queues, no scheduling, no orchestration DSL. ## In action Two agents spawned in one turn, working in parallel:

pi spawning two nano-team agents in parallel

## Tools Three of them, that's it: - `nano_agent_spawn(name, task?)` — run a team member. `task` overrides the YAML default. - `nano_agent_kill(name)` — abort a running agent. - `nano_agent_status(name?)` — markdown table of everyone, or one agent's full transcript. Issue several `spawn` calls in one turn and they go off in parallel. Chain them by feeding one's output into the next one's `task`. ## Adding an agent Drop a YAML file at `.pi/nano-team/team/.yaml`: ```yaml name: developer role: developer model: inception/mercury-2 instructions: | You write TypeScript that meets this project's standards. - strict mode; no `any`; functional style; immutable data - no comments unless the WHY is non-obvious task: | Implement the requested change end-to-end. State the file paths touched in the summary. ``` The required fields: - `name` — what you'll call them in `nano_agent_spawn` - `role` — one lowercased word (developer, reviewer, analyst…) - `model` — any model id pi knows about - `instructions` — system prompt for the subagent - `task` — default task; can be overridden per spawn Run `/reload` after editing. A few starter agents live in `examples/team/` if you want to copy from. ## Install From npm: ``` pi install npm:nano-team ``` Or from GitHub: ``` pi install git:github.com/daynin/nano-team ``` That writes to your global pi settings (`~/.pi/agent/settings.json`). Pass `-l` to install only for the current project. Other install sources work too — local path, https URL — see the [pi packages docs](https://github.com/badlogic/pi-mono/blob/main/packages/coding-agent/docs/packages.md) for the full list. Verify with `pi list`. Remove with `pi remove nano-team`. ## The faces Each agent gets its own color, picked from a 12-color palette so they don't blur into each other on the row. Five states (`idle`, `thinking`, `working`, `done`, `error`), four variants per state, animating frame by frame. Errors get crossed-out eyes and a frown. It's not load-bearing — it's just nicer to look at than a status bar. ## Stack - TypeScript strict, no build step. The extension loads as `.ts` source via jiti. - Deps: `@mariozechner/pi-coding-agent`, `yaml`, `typebox`. Nothing else. ## License MIT — see [LICENSE](LICENSE).