Octo is an open-source, AI-powered coding assistant designed to help developers with various coding tasks. It supports multiple LLM providers, including OpenAI and Anthropic, and allows users to switch models mid-conversation. Its primary benefit is enhancing productivity by automating tool calls and code edits, with optional custom-trained autofix models for improved reliability.

How It Works

Octo acts as an intermediary between the user and various LLM APIs. It intelligently manages multi-turn conversations and token usage, aiming to maximize the LLM's effectiveness. A key feature is its ability to integrate custom-trained "autofix" models that specifically handle failures in tool calls and code edits made by the primary LLM, offering a more robust and reliable coding assistance experience.

Quick Start & Requirements

• Install globally via npm: npm install --global octofriend
• Run from the command line: octofriend
• Supports any OpenAI-compatible or Anthropic-compatible LLM API.
• Configuration for local LLMs and MCP servers is available in ~/.config/octofriend/octofriend.json5.
• For verbose debugging output, set the OCTO_VERBOSE environment variable: OCTO_VERBOSE=1 octofriend.

Highlighted Details

• Seamlessly switch between LLM models (e.g., GPT-5, Claude 4) during a conversation.
• Optional, custom-trained autofix models for enhanced tool call and code edit handling.
• Zero telemetry, promoting user privacy.
• Supports integration with MCP servers for rich data access.
• Can be configured to use local LLMs via API base URLs.

Maintenance & Community

• Developed by synthetic-lab.
• No specific community links (Discord/Slack) or roadmap details are provided in the README.

Licensing & Compatibility

• The README does not explicitly state a license.

Limitations & Caveats

• The project's license is not specified, which may impact commercial use or closed-source integration.
• While it supports many LLMs, optimal performance might depend on the specific models used, especially for the autofix capabilities.