sruja
A CLI autonomous coding agent that writes code and verifies it with an independent grader to keep every edit aligned with your repo’s structure — an experimental tool for developers exploring self‑verifying code agents.
Tool overview
Adoption verdict: sruja is a very early‑stage open‑source project (22 stars) with no published benchmarks, tutorials, or in‑depth walkthroughs. It reads more like a proof‑of‑concept than a production‑ready tool, distinguished by its decision to completely decouple the actor from the evaluator.
What it does: Run from the CLI, it analyzes your repository’s file topology and runs a full autonomous cycle—comprehend, plan, execute, critique, replan—executing edits only if they pass independent deterministic checks for drift, style, task completion, and intent alignment. A passive MCP harness is included for tool integration.
Costs & fit: The agent itself is free and open‑source, but you must supply your own LLM API keys, so model usage costs apply (no official pricing; this is a conservative inference). It appeals to developers curious about autonomous coding loops and verification architecture, especially those comfortable with Rust. It is not a copilot‑style auto‑complete plugin, and it is not suitable for teams looking for a polished, one‑click coding assistant or for direct production use without heavy supervision.