Skip to content

CLI Reference

Initialize Fleet in the current repository.

Terminal window
fleet init

Creates .fleet/config.yml and the fleet/state orphan branch with an empty FLEET.md.


Start or resume the commander.

Terminal window
# Plan missions from a natural language goal
fleet command --plan "Add OAuth login, fix rate limiter, update docs"
# Load missions from a YAML plan file
fleet command --plan-file missions.yml
# Use a built-in mission template
fleet command --template test-coverage
# List available templates
fleet command --template list
# Resume an existing fleet (re-claim commander)
fleet command --resume
# Resume with graceful handoff support
fleet command --resume --handoff
FlagDescription
--plan <goal>Natural language goal for LLM decomposition
--plan-file <path>YAML file with pre-defined missions
--template <name>Use a built-in mission template; use list to see all
--resumeResume monitoring existing missions
--handoffEnable implicit commander handoff (any machine can resume)

See Mission Templates for the full list of built-in templates.


Run a mission template locally in sequence — no extra machines needed.

Terminal window
# Preview the full run experience without spawning real agents
fleet run --template drsti-dev-flow --simulate
# Run a built-in template on the current repo
fleet run --template drsti-dev-flow
# Override the agent adapter for all missions
fleet run --template test-coverage --agent codex
# Run against a specific directory
fleet run --template security-audit --cwd /path/to/project

fleet run is the single-machine shortcut for fleet command + fleet ship. It:

  1. Loads the named template and topologically sorts its missions by dependency.
  2. Checks pre-start gate conditions (if the drsti-dev-flow plugin is active).
  3. Spawns each mission’s agent adapter and waits for it to complete.
  4. Checks the merge gate between missions, with an interactive retry loop so you can update workstreams.json in place.
  5. Prints a branch summary when all missions are done.

--simulate mode runs the full terminal experience — headers, gate checks, mission progress, branch summary — without spawning any real agents or touching your repo. Use it to preview what a template will do before committing to a full run.

FlagDescription
--template <name>Built-in mission template to run (required)
--simulateSimulate the run without real agents (UAT / preview mode)
--cwd <path>Working directory (default: current directory)
--agent <adapter>Override the agent adapter for all missions

See Mission Templates for available templates.


Join the fleet as a worker ship.

Terminal window
fleet ship --join git@github.com:you/project.git

Clones the repo, reads FLEET.md, auto-assigns to the first ready mission, starts the coding agent, and begins heartbeat pushes.

FlagDescription
--join <repo>Git URL of the repository to work on

Display the mission board.

Terminal window
# One-shot display
fleet status
# Live-updating display
fleet status --watch
# Machine-readable JSON output
fleet status --json
FlagDescription
--watchRefresh the display every 30 seconds
--jsonOutput fleet state as JSON

Manually assign a mission to a ship.

Terminal window
fleet assign <mission-id> <ship-id>

Generate a codebase context file for ships.

Terminal window
# Static analysis (file tree, package.json, etc.)
fleet brief --generate
# LLM-powered analysis (richer context)
fleet brief --generate --llm

Creates FLEET_CONTEXT.md on the main branch. Ships read this file for codebase context before starting work.

FlagDescription
--generateGenerate the brief
--llmUse LLM for richer codebase analysis

View a ship’s MISSION.md log.

Terminal window
# One-shot display
fleet logs <ship-id>
# Follow mode (polls every 10s)
fleet logs <ship-id> --follow
FlagDescription
--followKeep polling for updates

Start the browser-based Fleet web dashboard.

Terminal window
fleet web

The dashboard shows mission board, ship health, merge queue, and mission logs in real time via Server-Sent Events. Accessible from any browser including mobile — the command prints both local and LAN URLs.

FlagDescription
-p, --port <port>HTTP port (default: 4000)
-H, --host <host>Bind address (default: 0.0.0.0)
--poll <ms>Poll interval in milliseconds (default: 15000)
--no-openDo not open browser automatically

Run a zero-friction simulated fleet — no repo, no agents, no network required.

Terminal window
fleet demo
# or
npx fleetspark demo

Simulates a complete fleet run with 4 missions, showing planning, ship assignment, parallel execution, dependency unblocking, and auto-merge. The ideal way to see Fleet in action before setting up a real fleet.


Generate a markdown summary of the current or most recent fleet run.

Terminal window
# Print to stdout
fleet report
# Machine-readable JSON
fleet report --json
# Write to a file
fleet report --output run-report.md

The report includes mission outcomes, timing, ships used, merge status, and an estimated time saved vs. sequential execution.

FlagDescription
--jsonOutput machine-readable JSON instead of markdown
--output <file>Write report to a file

Interactive terminal UI for monitoring the fleet in real-time.

Terminal window
fleet dashboard
# or
fleet dash

Features:

  • Mission board — color-coded status table with progress tracking
  • Ship health — live heartbeat monitoring with alive/stale/dead indicators
  • Merge queue — PR and CI status at a glance
  • Log viewer — drill into any mission’s steps and blockers
KeyAction
j/k or arrowsNavigate missions
TabSwitch between board and log views
qQuit

The dashboard auto-refreshes every 10 seconds.


Manage Fleet plugins — install governance layers, integrations, and extensions.

Terminal window
# Install a plugin and register it in .fleet/config.yml
fleet plugin install @fleetspark/plugin-drsti-dev-flow
# List registered plugins
fleet plugin list

Plugins are loaded automatically on every fleet ship --join and during merge operations. They can block missions from starting or block PRs from being created based on project-defined rules.

Installing drsti-dev-flow governance:

Terminal window
fleet plugin install @fleetspark/plugin-drsti-dev-flow

This registers the plugin in .fleet/config.yml:

plugins:
- name: "@fleetspark/plugin-drsti-dev-flow"

From that point on, Fleet enforces review gates before impl missions start and before PRs are created — based on the maturity level in each workstream’s workstreams.json entry.

See Ecosystem for the full plugin architecture.