s projects evolve, tools like Cursor, Claude Code, and Codex can start working from stale assumptions — old docs, deleted configs, outdated architecture decisions, or conflicting versions of the same project.

HARIKOS scans your repository, connects project facts back to evidence, tracks what’s current vs stale or conflicting, and exposes that context to your coding agents through MCP.

Built for developers and AI-heavy builders working on long-running projects where “just give the agent more context” stops being enough.

Core idea:
Repo → Evidence → Current Truth → Agent Context

The product is live and I’m currently looking for real users to test it on evolving codebases and tell me where it breaks.

Try HARIKOS:
https://harikos-ai.vercel.app

harikos

About