add agent teams implementation with tmux setup and Oishii prompt example
Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Shayan Rais
@@ -45,7 +45,7 @@ practice makes claude perfect
|
||||
| [**Scheduled Tasks**](https://code.claude.com/docs/en/scheduled-tasks) | `/loop`, cron tools | [](https://x.com/bcherny/status/2030193932404150413) [](implementation/claude-scheduled-tasks-implementation.md) Run prompts on a recurring schedule (up to 3 days), set one-time reminders, poll deployments and builds |
|
||||
| [**Voice Mode**](https://x.com/trq212/status/2028628570692890800)  | `/voice` | [](https://x.com/trq212/status/2028628570692890800) speak to prompt - /voice to activate|
|
||||
| [**Simplify & Batch**](https://x.com/bcherny/status/2027534984534544489) | `/simplify`, `/batch` | [](https://x.com/bcherny/status/2027534984534544489) Built-in skills for code quality and bulk operations — simplify refactors for reuse and efficiency, batch runs commands across files |
|
||||
| [**Agent Teams**](https://code.claude.com/docs/en/agent-teams)  | built-in (env var) | [](https://x.com/bcherny/status/2019472394696683904) Multiple agents working in parallel on the same codebase with shared task coordination |
|
||||
| [**Agent Teams**](https://code.claude.com/docs/en/agent-teams)  | built-in (env var) | [](https://x.com/bcherny/status/2019472394696683904) [](implementation/claude-agent-teams-implementation.md) Multiple agents working in parallel on the same codebase with shared task coordination |
|
||||
| [**Remote Control**](https://code.claude.com/docs/en/remote-control) | `/remote-control`, `/rc` | Continue local sessions from any device — phone, tablet, or browser · [Headless Mode](https://code.claude.com/docs/en/headless) |
|
||||
| [**Git Worktrees**](https://code.claude.com/docs/en/common-workflows) | built-in | [](https://x.com/bcherny/status/2025007393290272904) Isolated git branches for parallel development — each agent gets its own working copy |
|
||||
| [**Ralph Wiggum Loop**](https://github.com/anthropics/claude-code/tree/main/plugins/ralph-wiggum) | plugin | [](https://github.com/ghuntley/how-to-ralph-wiggum) [](https://github.com/shanraisshan/novel-llm-26) Autonomous development loop for long-running tasks — iterates until completion |
|
||||
|
||||
Binary file not shown.
|
After Width: | Height: | Size: 1.2 MiB |
@@ -0,0 +1,136 @@
|
||||
# Agent Teams Implementation
|
||||
|
||||
|
||||
|
||||
<table width="100%">
|
||||
<tr>
|
||||
<td><a href="../">← Back to Claude Code Best Practice</a></td>
|
||||
<td align="right"><img src="../!/claude-jumping.svg" alt="Claude" width="60" /></td>
|
||||
</tr>
|
||||
</table>
|
||||
|
||||
---
|
||||
|
||||
<a href="#asian-fine-dining-app"><img src="../!/tags/implemented-hd.svg" alt="Implemented"></a>
|
||||
|
||||
<p align="center">
|
||||
<img src="assets/impl-agent-teams.png" alt="Agent Teams in action — split pane mode with tmux" width="100%">
|
||||
</p>
|
||||
|
||||
Agent Teams spawn **multiple independent Claude Code sessions** that coordinate via a shared task list. Unlike subagents (isolated context forks within one session), each teammate gets its own full context window with CLAUDE.md, MCP servers, and skills loaded automatically.
|
||||
|
||||
---
|
||||
|
||||
## 
|
||||
|
||||
```bash
|
||||
tmux new -s oishii
|
||||
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 claude
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 
|
||||
|
||||
### 1. Install [iTerm2](https://iterm2.com/) and tmux
|
||||
|
||||
```bash
|
||||
brew install --cask iterm2
|
||||
brew install tmux
|
||||
```
|
||||
|
||||
### 2. Start iTerm2 → tmux → Claude
|
||||
|
||||
```bash
|
||||
tmux new -s dev
|
||||
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 claude
|
||||
```
|
||||
|
||||
### 3. Prompt with team structure
|
||||
|
||||
<a id="asian-fine-dining-app"></a>
|
||||
|
||||
```text
|
||||
Create an agent team to build "Oishii" — an Asian fine dining restaurant
|
||||
discovery app. The app lets users browse curated Asian fine dining restaurants,
|
||||
filter by cuisine (Japanese, Chinese, Thai, Korean, Vietnamese, Indian),
|
||||
view chef profiles, read tasting menu details, and make reservations.
|
||||
|
||||
Assign these teammates:
|
||||
|
||||
1. **Product Manager** — Define the MVP scope, write user stories for the
|
||||
core flows (browse restaurants, filter by cuisine, view restaurant detail,
|
||||
make a reservation). Create a product spec in docs/product-spec.md with
|
||||
acceptance criteria for each story. Coordinate with other teammates via
|
||||
the shared task list to unblock them.
|
||||
|
||||
2. **Senior Backend Engineer (API)** — Design and implement the REST API
|
||||
using Node.js/Express with TypeScript. Create the data models
|
||||
(Restaurant, Chef, Menu, Reservation), seed data with 12 curated Asian
|
||||
fine dining restaurants across 6 cuisines, and implement endpoints:
|
||||
GET /restaurants (with cuisine filter), GET /restaurants/:id,
|
||||
POST /reservations. Write the API in src/api/.
|
||||
|
||||
3. **Senior Backend Engineer (Database)** — Set up the SQLite database
|
||||
schema with Drizzle ORM. Create migrations, define relations between
|
||||
tables, and implement the repository layer in src/db/. Coordinate with
|
||||
the API engineer via tasks to agree on the data model interface.
|
||||
|
||||
4. **Frontend Engineer** — Build the React frontend with Tailwind CSS in
|
||||
src/app/. Create pages: restaurant listing with cuisine filter chips,
|
||||
restaurant detail with chef bio and tasting menu, and a reservation
|
||||
form modal. Use elegant typography and a dark theme befitting fine dining.
|
||||
Consume the API endpoints defined by the backend team.
|
||||
|
||||
5. **Deployment Engineer** — Set up the project scaffolding (package.json,
|
||||
tsconfig, vite config), Docker Compose for local dev (app + db),
|
||||
and a GitHub Actions CI pipeline (.github/workflows/ci.yml) that runs
|
||||
lint, typecheck, and tests. Write a README.md with setup instructions.
|
||||
|
||||
6. **QA Engineer** — Write end-to-end tests using Vitest for the API
|
||||
(test all endpoints, edge cases like invalid cuisine filter, double
|
||||
booking same time slot) and component tests for the React frontend.
|
||||
Create a test plan in docs/test-plan.md. Run the test suite and
|
||||
report failures to the team via tasks.
|
||||
|
||||
Each teammate should create tasks in the shared task list to coordinate
|
||||
dependencies (e.g., Backend DB must finish schema before API can seed data,
|
||||
API must be running before Frontend can integrate, QA needs both running
|
||||
before e2e tests). Start with Product Manager and Deployment Engineer
|
||||
in parallel, then fan out to the rest.
|
||||
```
|
||||
|
||||
### Team Coordination Flow
|
||||
|
||||
```
|
||||
┌─────────────────────────────────────────────────────────────┐
|
||||
│ LEAD (You) │
|
||||
│ "Create an agent team to build Oishii..." │
|
||||
└──────────────────────────┬──────────────────────────────────┘
|
||||
│ spawns team
|
||||
┌────────────┴────────────┐
|
||||
▼ ▼
|
||||
┌─────────────────┐ ┌──────────────────┐
|
||||
│ Product Manager │ │ Deployment Eng. │
|
||||
│ docs/spec.md │ │ scaffolding, │
|
||||
│ user stories │ │ Docker, CI │
|
||||
└────────┬────────┘ └────────┬─────────┘
|
||||
│ tasks: stories ready │ tasks: project ready
|
||||
▼ ▼
|
||||
┌──────────────────────────────────────────┐
|
||||
│ Shared Task List │
|
||||
│ ☐ Schema defined (DB → API) │
|
||||
│ ☐ Seed data loaded (API) │
|
||||
│ ☐ API endpoints live (API → FE) │
|
||||
│ ☐ UI components done (FE → QA) │
|
||||
└──────────────────────────────────────────┘
|
||||
│ │
|
||||
┌─────────┴──────┐ ┌────────┴────────┐
|
||||
▼ ▼ ▼ ▼
|
||||
┌──────────┐ ┌──────────┐ ┌────────────┐ ┌────────┐
|
||||
│ Backend │ │ Backend │ │ Frontend │ │ QA │
|
||||
│ (DB) │ │ (API) │ │ Engineer │ │Engineer│
|
||||
│ src/db/ │ │ src/api/ │ │ src/app/ │ │ tests/ │
|
||||
└──────────┘ └──────────┘ └────────────┘ └────────┘
|
||||
```
|
||||
|
||||
Reference in New Issue
Block a user