Apra Fleet from Apra Labs is an MCP server tool designed to coordinate AI agents across diverse machines and LLM providers. Built with JavaScript and running on Node.js 22+, it transforms agents like Claude Code, Gemini, and Copilot into orchestrators of a "fleet." Distributed via npm and standalone installers, Apra Fleet enables agents to dispatch commands to local or remote machines, integrating with issue trackers and supporting durable, multi-hour unattended workflows. It is a free program categorized as an mcps platform tool.

What does Apra Fleet connect to?

Apra Fleet functions as an MCP server and CLI, connecting to various MCP-capable AI agents, including Claude Code, Gemini, Codex, Copilot, OpenCode, Antigravity, ChatGPT, Goose, and VS Code. It registers and dispatches commands to "real machines," which can be local or remote via SSH. These machines run provider CLIs for the LLMs. Additionally, Apra Fleet integrates with issue trackers like beads from gascity, enhancing its utility for structured project management and autonomous software development.

Core features and tools

Apra Fleet exposes several callable tools to manage and interact with its fleet operations. These include apra-fleet install, apra-fleet start, apra-fleet workflow, apra-fleet consolelive loop, fleet manifest, apra-fleet update, and apra-fleet watch. It also incorporates a "PM skill" that activates /pm commands, specifically designed to support doer-reviewer sprints. The system primarily operates through conversational interaction with an MCP-capable agent, using plain language commands for setup and coordination.

Who is it for?

Apra Fleet targets developers, platform engineers, and organizations aiming to coordinate multiple AI agents across various machines and LLM providers. It suits teams requiring advanced scheduling, credential management, isolation, and observability for an agentic workforce. This system is ideal for those needing to manage durable, multi-hour, unattended workflows. Skip this if your operations do not involve orchestrating AI agents across multiple physical or virtual machines.

How to get started

The server authenticates using an API key passed as an environment variable, such as ANTIGRAVITY_API_KEY, and supports SAML for Single Sign-On.

Use cases

• Autonomous Software Development (fleet-sprint): Manages work from an issue tracker, dispatches across machines, uses AI for review, runs integration suites, and fixes bugs.
• Doer + Reviewer Loop: Establishes a fundamental pattern where one AI agent writes code and another reviews the changes.
• General Multi-Agent Orchestration: Provides scheduling, credential management, isolation, and observability for agentic workforces across diverse hardware and LLM providers.

Apra Fleet provides a specialized platform for orchestrating AI agents across various LLM providers and physical machines. Its support for MCP and A2A protocols, coupled with authentication options, makes it suitable for complex agentic workflows. The system's focus on conversational interaction simplifies managing multi-agent sprints and autonomous development cycles. Try running a fleet-sprint workflow to see how agents can plan, develop, and review code collaboratively.