matrixmapai/openabcode

What it solves

OpenABCode is a terminal-based coding agent that solves the problem of model selection by dynamically routing development tasks to the LLM best suited for that specific job. Instead of relying on a single model, it ensures that specialized tasks—like Google ecosystem development or code reviews—are handled by the models that perform best in those areas.

How it works

The agent uses a routing system based on a classifier and a set of configurable rules. It analyzes the user's prompt and project context (such as file extensions and project markers) to determine the best provider. By default, it routes Google ecosystem tasks to Gemini, code reviews and testing to ChatGPT, and general development to Claude.

Who it’s for

It is designed for developers who use multiple LLM providers and want to optimize the quality of their code generation and review without manually switching models for every task.

Highlights

• Dynamic Routing: Automatically assigns tasks to the most capable model based on the system's classification logic.
• Customizable Rules: Users can override routing logic via global or project-level settings to define their own keywords, file extensions, and project markers.
• Multi-Provider Support: Integrates with OpenAI, Anthropic, and Google via a unified API.
• Containerization Options: Supports isolation patterns using Docker, OpenShell, or the Gondolin extension to secure tool execution.
• Interactive CLI: Provides a terminal UI for managing sessions, selecting models, and auditing routing decisions.