Fleet 2.9.1 is out.See what's new →
FleetFleet
MCP solutions

26 Fleet MCP tool solutions for any MCP client

Fleet turns its Model Context Protocol integration into concrete developer workflows: local agent control, pipeline gates, Fabric Event Bus handoffs, Brain Daemon insight, task routing, issue watches, and config management from Claude Code, Codex, OpenCode, or any other MCP client.

Read MCP setupInstall Fleet

Self-hosted

Claude Code, Codex, OpenCode, and other MCP clients talk to `fleet mcp serve` over stdio, or to hosted Streamable HTTP. No extra network port for the local transport.

Audit trail

Fabric events, pipeline approvals, task updates, and agent activity remain queryable after the prompt ends.

Loop engineering

Each tool turns a repeated management prompt into a reusable local workflow.

Choose the tool-sized entry point

Every page maps one MCP tool to the developer problem it solves, an example prompt, and the guardrail Fleet keeps around the action.

Fleet & Status Tools

Inspect health, diagnostics, logs, dashboards, and Brain insights.

fleet_status

check Fleet health

Get agent counts, pending deliveries, and Brain alerts.

fleet_doctor

diagnose the Fleet host

Check tmux, SQLite, data paths, sockets, and GitHub CLI.

fleet_log

query the audit timeline

Read the unified Fabric and agent activity timeline.

fleet_dashboard

render an executive dashboard

Generate a CEO, CTO, CPO, CMO, or general operating view.

fleet_brain_insights

inspect agent quality and risk

Return evaluation, risk, thoughts, and recommendations.

Agent Tools

Create, inspect, start, stop, and remove Fleet-managed agents.

fleet_agent_list

list agents

List agents with status, vendor, model, and mode.

fleet_agent_start

start an agent

Start an enabled, stopped agent by fuzzy name.

fleet_agent_stop

stop an agent

Stop a running agent by fuzzy name.

fleet_agent_create

create an agent

Create a configured agent instance.

fleet_agent_delete

delete an agent

Delete an agent, optionally stopping it first.

fleet_agent_logs

read agent logs

Return raw agent log lines.

fleet_agent_output

read parsed run output

Return run times, exit codes, durations, PRs, and commits.

Fabric Tools

Publish coordination events and inspect pending inbox deliveries.

fleet_fabric_publish

publish a Fabric event

Publish a task, handoff, blocker, decision, update, request, or review event.

fleet_fabric_inbox

read a Fabric inbox

List deliveries by agent and status.

Watch Tools

Notify an agent when a watched GitHub issue closes.

fleet_watch_create

create an issue watch

Register a project-scoped issue-close watch.

fleet_watch_list

list issue watches

List active watches, optionally by agent.

Task & Context Tools

Assign work and inspect or update Fabric task state.

fleet_agent_context

read an agent's context

Return the agent config and pending inbox.

fleet_task_assign

assign and start a task

Set FLEET_TASK_TITLE and start the selected agent immediately.

fleet_task_get

get a task

Return title, description, status, priority, and assignee.

fleet_task_list

list tasks

List tasks, optionally filtered by status.

fleet_task_update

update a task

Change task status or assignee.

Configuration Tools

Read and update local Fleet settings.

fleet_config_list

list settings

List configuration as key/value pairs.

fleet_config_get

get a setting

Read one setting by key.

fleet_config_set

set a setting

Write one setting key and value.

Workflow Tools

Inspect and update saved control-plane workflow execution settings.

fleet_workflow_get

get a saved workflow

Read a saved control-plane workflow by its opaque ID.

fleet_workflow_update

update workflow execution settings

Update a saved workflow's runner and model without replacing its definition.

Connect any MCP client to Fleet

Install the local binary, add `.mcp.json` or the hosted Streamable HTTP URL, and give Claude Code, Codex, OpenCode, or any other MCP client local tools for managing your agent fleet.

See how it worksInstall Fleet