A multi-provider AI coding agent with the persona of a Tech-Priest

Binharic is a terminal-based AI coding assistant (a coding agent) similar to OpenAI's Codex, Google's Gemini CLI, and Anthropic's Claude Code—but with the personality of a devout Tech-Priest of the Adeptus Mechanicus. Binharic is written in TypeScript and uses the AI SDK framework for most of its underlying agentic logic (like tool calling and workflow management). Additionally, its architecture follows the recommendations mentioned in the building effective agents article from Anthropic, to a good degree.

Binharic's development started as a personal project to learn more about building a terminal-based coding agent. However, the project has grown somewhat into a full-fledged coding assistant with a lot of features like the ability to analyze projects, run tests, find bugs, and perform code review.

Features

• Can use models from OpenAI, Google, Anthropic, and Ollama
• Is fully customizable and extendable (system prompt and adding new tools)
• Comes with a built-in keyword-based retrieval-augmented generation (RAG) pipeline
• Comes with a large set of built-in tools (like for reading files and running Bash commands)
• Can use external tools via the Model Context Protocol (MCP)
• Comes with predefined workflows for common software development tasks (like debugging and code review)

See the ROADMAP.md for the list of implemented and planned features.

[!IMPORTANT] Binharic is in early development, so bugs and breaking changes are expected. Please use the issues page to report bugs or request features.

Getting Started

You can follow the instructions below to install and use Binharic in your terminal.

Installation

npm install -g @cogitator/binharic-cli

Running in the Terminal

# Make sure API keys are available in the environment
export OPENAI_API_KEY=<your-openai-api-key>
export ANTHROPIC_API_KEY=<your-anthropic-api-key>
export GOOGLE_API_KEY=<your-google-api-key>

# Start Binharic in the terminal
binharic

[!NOTE] The performance of a coding agent like Binharic, to a great extent, depends on the model it uses. So, it's recommended to use state-of-the-art models (like Claude Sonnet 4.5, GPT-5, and Gemini 2.5 Pro) for the best results.

Running in a Container

Alternatively, you can run Binharic in a container:

# API keys should be available in the environment already
docker run -it --rm \
  -v ${PWD}:/workspace \
  -e OPENAI_API_KEY \
  -e ANTHROPIC_API_KEY \
  -e GOOGLE_API_KEY \
  ghcr.io/cogitatortech/binharic-cli:<version>

<version> should be replaced with the version of the Binharic (like 0.1.0-alpha.6) or latest. Use latest if you want to use the latest (development) version of Binharic.

Configuration

You can configure Binharic by editing the ~/.config/binharic/config.json5 file. Additionally, Binharic supports environment variables for configuration. See the docs (TO BE ADDED) for more information.

Documentation

See the docs for more information on how to use the Binharic coding agent.

Contributing

See CONTRIBUTING.md for details on how to make a contribution.

License

This project is licensed under the MIT License (see LICENSE).

Acknowledgements

• The logo is from SVG Repo with some modifications.