A coding agent that runs on your machine.
A local coding and integration agent. The browser is a thin client; repository indexing, file edits, and every conversation happen inside a sealed engine on your own hardware. By default, inference, indexing, and conversations stay on your machine and nothing is sent to OpenEng; a model you point at an external server, and any MCP server you attach, receive what you send them.
In the app
The Chat web app is live, but its engine hasn’t been published yet — there’s nothing to install today. The one-line installer will appear here when it ships; your OpenEng account already works across the suite.
Chat features
Hand inference to a local Ollama or LM Studio server, or to any OpenAI-compatible server you point it at, which then receives what you send it; built-in llama.cpp and mistral.rs runtimes, with downloads managed by the engine, are planned for the release build.
A tree-sitter repo map lets the agent navigate and edit real codebases by structure rather than by guessing at file names.
Attach Terminal tabs, Kubernetes clusters, and Git repos to a chat group, and the agent works them through their own local engines over the same sealed loopback channel — one conversation can run a command, read pod logs, and check a repo’s status.
Group conversations under a shared name and system prompt, and point each group at the connections it is allowed to touch.
Attach Model Context Protocol servers to give the agent extra tools, hosted by the engine alongside its built-in ones.
The UI reaches the engine on 127.0.0.1 over a channel sealed frame by frame. Projects, conversations, and model files are stored on your machine, never on an OpenEng server.
The agent retrieves context, plans, edits files, and runs your build or tests to verify — every plan, tool call, diff, and verify result streams into the transcript, with auto-apply when you want it.
Each project’s Analysis tab summarizes its local index: files, symbols, edges, and languages, the most-referenced hotspots (PageRank), and entry points.
Attach images, video frames, PDFs, or code files — up to five per message — for the model to read. On-device speech-to-text dictation is planned for the release build.
Chat is a thin browser client of a local engine you run yourself. Your models, conversations, and code stay on your hardware: the engine listens only on 127.0.0.1, and the page seals every frame of its session with the engine.
The Chat engine hasn’t been published yet, so there is nothing to install today — its one-line installer will appear here and on the download page when it ships. Every engine and platform
The rest of the suite
Every OpenEng app is free — your account works across Terminal, Kubernetes, Data, Git, and Chat.
One platform, two pillars
The engineering apps are one half of OpenEng. The other is OpenEng AI, an agentic ecosystem for open-weight model execution: it downloads, orchestrates, and runs open-weight AI models on your own machine.
Explore OpenEng AI
