Claude DevFleet Multi-Agent Orchestration
When to Use
Use this skill when you need to dispatch multiple Claude Code agents to work on coding tasks in parallel. Each agent runs in an isolated git worktree with full tooling.
Setup
The DevFleet server is a separate project, not bundled with ECC. Install and
run it from its repository first: https://github.com/LEC-AI/claude-devfleet
Then connect the running instance via MCP:
claude mcp add devfleet --transport http http://localhost:18801/mcp
Before first use, verify the process listening on port 18801 is the DevFleet
binary you installed (see SECURITY.md on localhost MCP servers).
How It Works
User → "Build a REST API with auth and tests"
↓
plan_project(prompt) → project_id + mission DAG
↓
Show plan to user → get approval
↓
dispatch_mission(M1) → Agent 1 spawns in worktree
↓
M1 completes → auto-merge → auto-dispatch M2 (depends_on M1)
↓
M2 completes → auto-merge
↓
get_report(M2) → files_changed, what_done, errors, next_steps
↓
Report back to user
Tools
| Tool |
Purpose |
plan_project(prompt) |
AI breaks a description into a project with chained missions |
create_project(name, path?, description?) |
Create a project manually, returns project_id |
create_mission(project_id, title, prompt, depends_on?, auto_dispatch?) |
Add a mission. depends_on is a list of mission ID strings (e.g., ["abc-123"]). Set auto_dispatch=true to auto-start when deps are met. |
dispatch_mission(mission_id, model?, max_turns?) |
Start an agent on a mission |
cancel_mission(mission_id) |
Stop a running agent |
wait_for_mission(mission_id, timeout_seconds?) |
Block until a mission completes (see note below) |
get_mission_status(mission_id) |
Check mission progress without blocking |
get_report(mission_id) |
Read structured report (files changed, tested, errors, next steps) |
get_dashboard() |
System overview: running agents, stats, recent activity |
list_projects() |
Browse all projects |
list_missions(project_id, status?) |
List missions in a project |
Note on wait_for_mission: This blocks the conversation for up to timeout_seconds (default 600). For long-running missions, prefer polling with get_mission_status every 30–60 seconds instead, so the user sees progress updates.
Workflow: Plan → Dispatch → Monitor → Report
- Plan: Call
plan_project(prompt="...") → returns project_id + list of missions with depends_on chains and auto_dispatch=true.
- Show plan: Present mission titles, types, and dependency chain to the user.
- Dispatch: Call
dispatch_mission(mission_id=<first_mission_id>) on the root mission (empty depends_on). Remaining missions auto-dispatch as their dependencies complete (because plan_project sets auto_dispatch=true on them).
- Monitor: Call
get_mission_status(mission_id=...) or get_dashboard() to check progress.
- Report: Call
get_report(mission_id=...) when missions complete. Share highlights with the user.
Concurrency
DevFleet runs up to 3 concurrent agents by default (configurable via DEVFLEET_MAX_AGENTS). When all slots are full, missions with auto_dispatch=true queue in the mission watcher and dispatch automatically as slots free up. Check get_dashboard() for current slot usage.
Examples
Full auto: plan and launch
plan_project(prompt="...") → shows plan with missions and dependencies.- Dispatch the first mission (the one with empty
depends_on).
- Remaining missions auto-dispatch as dependencies resolve (they have
auto_dispatch=true).
- Report back with project ID and mission count so the user knows what was launched.
- Poll with
get_mission_status or get_dashboard() periodically until all missions reach a terminal state (completed, failed, or cancelled).
get_report(mission_id=...) for each terminal mission — summarize successes and call out failures with errors and next steps.
Manual: step-by-step control
create_project(name="My Project") → returns project_id.create_mission(project_id=project_id, title="...", prompt="...", auto_dispatch=true) for the first (root) mission → capture root_mission_id.
create_mission(project_id=project_id, title="...", prompt="...", auto_dispatch=true, depends_on=["<root_mission_id>"]) for each subsequent task.dispatch_mission(mission_id=...) on the first mission to start the chain.get_report(mission_id=...) when done.
Sequential with review
create_project(name="...") → get project_id.create_mission(project_id=project_id, title="Implement feature", prompt="...") → get impl_mission_id.dispatch_mission(mission_id=impl_mission_id), then poll with get_mission_status until complete.get_report(mission_id=impl_mission_id) to review results.create_mission(project_id=project_id, title="Review", prompt="...", depends_on=[impl_mission_id], auto_dispatch=true) — auto-starts since the dependency is already met.
Guidelines
- Always confirm the plan with the user before dispatching, unless they said to go ahead.
- Include mission titles and IDs when reporting status.
- If a mission fails, read its report before retrying.
- Check
get_dashboard() for agent slot availability before bulk dispatching.
- Mission dependencies form a DAG — do not create circular dependencies.
- Each agent runs in an isolated git worktree and auto-merges on completion. If a merge conflict occurs, the changes remain on the agent's worktree branch for manual resolution.
- When manually creating missions, always set
auto_dispatch=true if you want them to trigger automatically when dependencies complete. Without this flag, missions stay in draft status.
1---2name: claude-devfleet3description: Orchestrate multi-agent coding tasks via Claude DevFleet — plan projects, dispatch parallel agents in isolated worktrees, monitor progress, and read structured reports. Use when dispatching parallel coding agents across isolated worktrees and tracking their reports.4---56# Claude DevFleet Multi-Agent Orchestration78## When to Use910Use this skill when you need to dispatch multiple Claude Code agents to work on coding tasks in parallel. Each agent runs in an isolated git worktree with full tooling.1112## Setup1314The DevFleet server is a separate project, not bundled with ECC. Install and15run it from its repository first: <https://github.com/LEC-AI/claude-devfleet>1617Then connect the running instance via MCP:18```bash19claude mcp add devfleet --transport http http://localhost:18801/mcp20```2122Before first use, verify the process listening on port 18801 is the DevFleet23binary you installed (see SECURITY.md on localhost MCP servers).2425## How It Works2627```28User → "Build a REST API with auth and tests"29 ↓30plan_project(prompt) → project_id + mission DAG31 ↓32Show plan to user → get approval33 ↓34dispatch_mission(M1) → Agent 1 spawns in worktree35 ↓36M1 completes → auto-merge → auto-dispatch M2 (depends_on M1)37 ↓38M2 completes → auto-merge39 ↓40get_report(M2) → files_changed, what_done, errors, next_steps41 ↓42Report back to user43```4445### Tools4647| Tool | Purpose |48|------|---------|49| `plan_project(prompt)` | AI breaks a description into a project with chained missions |50| `create_project(name, path?, description?)` | Create a project manually, returns `project_id` |51| `create_mission(project_id, title, prompt, depends_on?, auto_dispatch?)` | Add a mission. `depends_on` is a list of mission ID strings (e.g., `["abc-123"]`). Set `auto_dispatch=true` to auto-start when deps are met. |52| `dispatch_mission(mission_id, model?, max_turns?)` | Start an agent on a mission |53| `cancel_mission(mission_id)` | Stop a running agent |54| `wait_for_mission(mission_id, timeout_seconds?)` | Block until a mission completes (see note below) |55| `get_mission_status(mission_id)` | Check mission progress without blocking |56| `get_report(mission_id)` | Read structured report (files changed, tested, errors, next steps) |57| `get_dashboard()` | System overview: running agents, stats, recent activity |58| `list_projects()` | Browse all projects |59| `list_missions(project_id, status?)` | List missions in a project |6061> **Note on `wait_for_mission`:** This blocks the conversation for up to `timeout_seconds` (default 600). For long-running missions, prefer polling with `get_mission_status` every 30–60 seconds instead, so the user sees progress updates.6263### Workflow: Plan → Dispatch → Monitor → Report64651. **Plan**: Call `plan_project(prompt="...")` → returns `project_id` + list of missions with `depends_on` chains and `auto_dispatch=true`.662. **Show plan**: Present mission titles, types, and dependency chain to the user.673. **Dispatch**: Call `dispatch_mission(mission_id=<first_mission_id>)` on the root mission (empty `depends_on`). Remaining missions auto-dispatch as their dependencies complete (because `plan_project` sets `auto_dispatch=true` on them).684. **Monitor**: Call `get_mission_status(mission_id=...)` or `get_dashboard()` to check progress.695. **Report**: Call `get_report(mission_id=...)` when missions complete. Share highlights with the user.7071### Concurrency7273DevFleet runs up to 3 concurrent agents by default (configurable via `DEVFLEET_MAX_AGENTS`). When all slots are full, missions with `auto_dispatch=true` queue in the mission watcher and dispatch automatically as slots free up. Check `get_dashboard()` for current slot usage.7475## Examples7677### Full auto: plan and launch78791. `plan_project(prompt="...")` → shows plan with missions and dependencies.802. Dispatch the first mission (the one with empty `depends_on`).813. Remaining missions auto-dispatch as dependencies resolve (they have `auto_dispatch=true`).824. Report back with project ID and mission count so the user knows what was launched.835. Poll with `get_mission_status` or `get_dashboard()` periodically until all missions reach a terminal state (`completed`, `failed`, or `cancelled`).846. `get_report(mission_id=...)` for each terminal mission — summarize successes and call out failures with errors and next steps.8586### Manual: step-by-step control87881. `create_project(name="My Project")` → returns `project_id`.892. `create_mission(project_id=project_id, title="...", prompt="...", auto_dispatch=true)` for the first (root) mission → capture `root_mission_id`.90 `create_mission(project_id=project_id, title="...", prompt="...", auto_dispatch=true, depends_on=["<root_mission_id>"])` for each subsequent task.913. `dispatch_mission(mission_id=...)` on the first mission to start the chain.924. `get_report(mission_id=...)` when done.9394### Sequential with review95961. `create_project(name="...")` → get `project_id`.972. `create_mission(project_id=project_id, title="Implement feature", prompt="...")` → get `impl_mission_id`.983. `dispatch_mission(mission_id=impl_mission_id)`, then poll with `get_mission_status` until complete.994. `get_report(mission_id=impl_mission_id)` to review results.1005. `create_mission(project_id=project_id, title="Review", prompt="...", depends_on=[impl_mission_id], auto_dispatch=true)` — auto-starts since the dependency is already met.101102## Guidelines103104- Always confirm the plan with the user before dispatching, unless they said to go ahead.105- Include mission titles and IDs when reporting status.106- If a mission fails, read its report before retrying.107- Check `get_dashboard()` for agent slot availability before bulk dispatching.108- Mission dependencies form a DAG — do not create circular dependencies.109- Each agent runs in an isolated git worktree and auto-merges on completion. If a merge conflict occurs, the changes remain on the agent's worktree branch for manual resolution.110- When manually creating missions, always set `auto_dispatch=true` if you want them to trigger automatically when dependencies complete. Without this flag, missions stay in `draft` status.
JunMystery