OpenMono: AI Coding Agent

OpenMonoAgent.ai is a terminal-native coding agent for local LLM use, built on C/.NET. It is described as open source, free forever, and intended for use on the user’s own machine rather than through cloud services. The site frames it as infrastructure to own, with unlimited tokens and no API keys.

Its feature set centers on local coding assistance. The agent auto-detects hardware and loads an appropriate model, with one path for GPU use and another for CPU fallback. It includes a full-screen ANSI terminal interface with streaming tokens, a live tok/s meter, context-window usage, auto-compaction, and visual feedback while it thinks. The agent runs in Docker, bind-mounts the user’s project, and uses permission gates for destructive operations. It also advertises 20 tools plus Model Context Protocol support, including file I/O, shell access, search, web fetch, LSP features, patches, sub-agents, and plan mode.

The product is built for .NET developers and includes Roslyn-based compiler intelligence for type hierarchies, call graphs, and cross-assembly references. It also provides LSP support for C and TypeScript, with hover, go-to-definition, and reference lookup. Additional features listed on the site include playbooks for typed, composable, stateful workflow automation, persistent sessions with JSONL transcripts, cross-session memory, snapshot-based file undo, and a dual-box mode that runs the agent on a laptop while inference runs on a home GPU rig behind NAT or CGNAT.

Installation is presented as a single command, with the installer detecting hardware, building llama.cpp, downloading the model, and wiring up Docker. The site also says the tool is free to use indefinitely, with no per-token billing, rate limits, or surprise invoices. OpenMonoAgent.ai is open source and the page includes a GitHub star prompt and a mobile app label marked beta.