Run AI agents locally. Build workflows. Own your data.
Intent becomes the interface 🎯
Website · Downloads · Getting Started · YouTube · Issues
dotPilot is an open source desktop app for running AI agents locally on your machine.
- 🤖 Run multiple agents — Launch and manage several AI agents at once
- 🔄 Build workflows — Create multi-agent pipelines with Microsoft Agent Framework
- 🔌 Any provider — Codex, Claude Code, GitHub Copilot, Gemini, or local models
- 🏠 100% local — Your data stays on your device, no cloud required
- 💬 Natural language — Create agents by describing what you need
Built with C#, .NET 10, and Uno Platform.
| Platform | Architecture | Link |
|---|---|---|
| 🍎 macOS | Apple Silicon | Download .dmg |
| 🪟 Windows | x64 | Download .exe |
| 🐧 Linux | x64 | Download .snap |
| Provider | Type | Status |
|---|---|---|
| Codex CLI | OpenAI coding agent | ✅ |
| Claude Code | Anthropic assistant | ✅ |
| GitHub Copilot | Microsoft AI | ✅ |
| Gemini | Google AI | ✅ |
| OpenAI API | Direct API | ✅ |
| Azure OpenAI | Enterprise API | ✅ |
| LLamaSharp | Local models | ✅ |
| ONNX Runtime | Local inference | ✅ |
- Download the latest release for your platform from Releases
- Install and launch dotPilot
- Configure your preferred AI provider in Settings
- Create your first agent and start chatting
git clone https://github.com/managedcode/dotPilot.git
cd dotPilot
dotnet build DotPilot.slnx
dotnet run --project DotPilot/DotPilot.csproj -f net10.0-desktopRequirements: .NET SDK 10.0.103, Uno.Sdk 6.5.31
Launch as many AI agents as you need, all running in parallel on your local machine. Each agent operates with its own isolated context, memory, and tool set. The Fleet Board gives you a real-time dashboard to monitor all active sessions — see which agents are working, what they're doing, and their current status. Switch between agents instantly, pause or resume sessions, and manage your entire agent fleet from one unified interface.
dotPilot integrates with Microsoft Agent Framework to enable sophisticated multi-agent orchestration. Design workflows where agents collaborate — one agent researches, another analyzes, a third generates code. Use sequential pipelines for step-by-step tasks, parallel execution for independent workloads, or handoff patterns where agents pass context to each other. All workflows support streaming output, checkpoints for long-running tasks, and human-in-the-loop approvals when you need to stay in control.
Freedom to choose your AI backend. Connect to Codex CLI for OpenAI's coding agent, Claude Code for Anthropic's assistant, GitHub Copilot for Microsoft's AI pair programmer, or Gemini for Google's models. Need direct API access? Use OpenAI API or Azure OpenAI for enterprise scenarios. Want full privacy? Run models entirely on your hardware with LLamaSharp or ONNX Runtime — zero data leaves your machine. Mix and match providers across different agents based on what each task needs.
Stop configuring, start describing. Tell dotPilot what kind of agent you need in plain language: "I need an agent that reviews pull requests and suggests improvements" or "Create a research agent that summarizes technical papers". The system generates the complete agent profile — system prompt, tool configuration, provider selection, and behavioral settings. Refine with follow-up instructions or dive into manual config when you need fine-grained control. Intent becomes the interface.
Every agent action is traceable. dotPilot integrates OpenTelemetry to capture detailed telemetry for each session — every prompt sent, every response received, every tool invocation, every workflow step. Visualize agent reasoning flows, identify performance bottlenecks, debug unexpected behaviors. Export traces to your preferred observability backend or analyze locally. When something goes wrong, you'll know exactly where and why.
Your data never leaves your device unless you explicitly choose a cloud provider. All session history, agent configurations, and conversation transcripts are stored locally in SQLite. No telemetry sent to external servers. No account required to use local models. Run completely air-gapped if needed. You own your data, your workflows, and your AI infrastructure.
| Language | C# |
| Runtime | .NET 10 |
| UI | Uno Platform |
| Orchestration | Microsoft Agent Framework |
| Database | SQLite + EF Core |
| Observability | OpenTelemetry |
We're developing dotPilot in the open.
- 📺 YouTube — @ManagedCode
- 💬 Issues — GitHub Issues
- 🌐 Website — dotpilot.managed-code.com
Open source. See LICENSE for details.
Made by ManagedCode