OpenMonoAgent.ai provides a free, local-first AI coding agent designed to democratize access to powerful AI tools. It addresses the high costs and data privacy concerns associated with cloud-based AI agents by running entirely on user hardware, offering unlimited tokens without subscriptions or data leakage. This makes it ideal for developers, researchers, and power users seeking complete control over their AI development environment.

How It Works

The project employs a .NET 10 CLI that interfaces with a bundled llama.cpp inference server, all orchestrated within Docker containers. This local-first architecture ensures all processing, including model inference and code analysis, remains on the user's machine. The approach prioritizes data privacy and eliminates per-token costs, positioning AI tooling as owned infrastructure rather than a recurring subscription service.

Quick Start & Requirements

Install via: bash <(curl -fsSL https://raw.githubusercontent.com/StartupHakk/OpenMonoAgent.ai/refs/heads/main/get-openmono.sh)

Run using openmono agent (TUI) or openmono agent --classic (CLI). Requires Ubuntu 26.04 LTS (recommended) or 25.10. Hardware specifications include a minimum of 24 GB VRAM for GPU inference (~45 tok/s) or 24 GB RAM for CPU inference (~20 tok/s).

• Docs: docs/
• Roadmap: Roadmap
• Contributing: Contributing

Highlighted Details

• Bundled Inference: Zero-configuration llama.cpp server with auto-detection for GPU (Qwen3.6-27B) and CPU (Qwen3.6-35B) models.
• Agentic Loop: Features a robust agentic loop with a maximum of 25 iterations, doom-loop detection, and context window management.
• Extensive Tooling: Integrates 20 built-in tools within a 12-step processing pipeline, including read-only parallel execution.
• Specialist Sub-Agents: Employs 5 distinct sub-agents (Explore, Plan, Coder, Verify, general-purpose) with locked tool sets and turn budgets for focused tasks.
• Docker Sandboxing: Provides secure execution within a Docker container, mounting the project directory as /workspace for controlled file access.
• Deep Code Intelligence: Offers native Roslyn analysis for C#, LSP support for multiple languages (Python, Go, Rust), and optional integration with graphify and code-review-graph.
• Playbooks: Supports composable YAML-defined workflows with typed parameters and conditional logic.
• Provider Flexibility: Defaults to local llama.cpp, with OpenAI, Anthropic, and Ollama support marked as Work In Progress (WIP).

Maintenance & Community

Contributions are actively welcomed for new tools, providers, LSP servers, playbooks, bug fixes, and documentation. The project is developed by StartupHakk and built on C#/.NET. Community interaction details beyond the contributing guide are not explicitly detailed in the README.

Licensing & Compatibility

Licensed under the GNU AFFERO GENERAL PUBLIC LICENSE v3.0. This strong copyleft license requires derivative works to be made available under the same terms, potentially impacting integration with proprietary software.

Limitations & Caveats

The project is currently in Public Beta. A known limitation is the fixed maximum of 25 iterations per agent turn, with ongoing work to address potential performance degradation if this limit is significantly increased. Support for non-default inference providers (OpenAI, Anthropic, Ollama) is still under development.