apra-fleet is an innovative, open-source Model Context Protocol (MCP) server designed to maximize the utility of your existing hardware. It seamlessly turns the machines you already own into a coordinated, high-performing fleet of autonomous AI agents.

The platform empowers developers to orchestrate complex operations by mixing and matching diverse AI models—including Claude, Antigravity, Codex, and Copilot—as well as running local models through OpenCode. This hybrid approach ensures you can leverage the best model for every specific sub-task.

Key capabilities of apra-fleet include:

Cost-Tier Routing: Intelligently routes workloads based on cost tiers, helping you keep operational expenses optimized.
Durable & Observable Workflows: Built to handle resilient, multi-hour workflows with complete observability, ensuring long-running processes execute reliably.
Autonomous Building: The system is designed to autonomously build itself, reducing manual overhead and configuration bottlenecks.

Pros of apra-fleet

Fully open-source and designed to run on your own hardware
Supports a wide mix of models including Claude, Codex, Copilot, and local LLMs via OpenCode
Optimizes budget with smart routing based on cost tiers
Supports long-running, durable, and highly observable workflows
Features autonomous self-building capabilities

Cons of apra-fleet

Requires owning and maintaining your own machine infrastructure
Requires technical familiarity with MCP (Model Context Protocol) servers and developer tools
No pre-configured managed cloud hosting option is provided out of the box

Frequently Asked Questions

What is apra-fleet?

apra-fleet is an open-source MCP (Model Context Protocol) server that allows you to turn your existing machines and hardware into a coordinated fleet of AI agents.

Which models does apra-fleet support?

You can mix and route tasks between Claude, Antigravity, Codex, Copilot, and various local models via OpenCode.

How does work routing function in apra-fleet?

Work is automatically routed across your model fleet based on specified cost tiers, ensuring computationally expensive tasks are sent to high-tier models only when necessary.

Can it run tasks that take a long time to complete?

Yes. apra-fleet is built specifically to support durable, multi-hour, and observable workflows, ensuring reliability and monitoring throughout long-running tasks.