I wanted a Cursor-style agent that runs on my own model, so I built one

Description and use

OpenPilot is a desktop app that lets users run a Cursor‑style coding agent on any OpenAI‑compatible model, with full control over file access and command execution.

Problem

The author wanted a Cursor-like coding agent that could run on a locally hosted or self‑hosted model, but no open‑source tool met that requirement.

Outcome

OpenPilot enables users to run a Cursor‑style coding agent on any OpenAI‑compatible model, with full control over file access, command execution, and optional web search, while providing persistent memory and token usage metrics.

Workflow

• 1. Download the latest build from GitHub Releases or run from source.
• 2. On first launch, connect a model by providing base URL, API key, and model name.
• 3. Choose a workspace folder that the agent may modify.
• 4. Optionally add a TinyFish key for web search.
• 5. Describe a task in the chat (e.g., "add a README and a failing test for this function").
• 6. Observe tool cards and approve or deny actions as prompted.

Evidence

I looked around for an open-source tool that worked the way I had in mind and couldn't find one that fit. So I built it.

Source

OpenPilot is a desktop app, built with Electron. You give it a model, a folder, and a task.

Source

You stay in control. Tool calls show up as cards in the chat, and sensitive actions wait for you to approve or deny them.

Source

Windows is the main target right now. macOS builds exist but are early and unsigned. There's no Linux build yet.

Source