Autonoma

About

Autonoma is a Python framework in which a multi-agent system autonomously modifies, analyzes, and tests codebases through an event-driven loop. A PlannerAgent splits a query into tasks (up to 10 by default), CoderAgent generates or refactors code with AST manipulation and static analysis, and a Tester generates and runs unittest suites, feeding failures back for revision. Agents operate asynchronously with Pydantic-validated data structures, and the LLM interface is pluggable: any provider with a generate(prompt) method works, so backend choice is fully decoupled. Tasks can specify file paths and complexity estimates, and the architecture supports custom agents and task types. The project is an early-stage MIT-licensed Python package installed via pip, best viewed as a reference design for plan-code-test loops rather than a production harness.