A CLI-first autonomous coding agent where the actor never grades itself. Every edit is grounded in your repo's real topology and verified by an independent deterministic grader (drift, lint, verify-task, intent) before it ships. Closed-loop: comprehend → plan → execute → critique → replan. Also a passive MCP harness.
Latest release sruja-v0.70.0 — sruja: v0.70.0 · 21 Jul 2026
4 files for Cursor, GitHub Copilot, Codex, OpenCode and Claude Code: cursorrules, sruja copilot-instructions.md, sruja AGENTS.md, sruja CLAUDE.md — 7,132 tokens loaded in every session.
.github/copilot-instructions.md A 86 tok AGENTS.md A 6,175 tok CLAUDE.md A 588 tok .cursorrules A 283 tok These files are sruja-ai/sruja's own configuration — they tell Cursor, GitHub Copilot, Codex, OpenCode and Claude Code how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.