AccInt

AccInt (Accreted Intelligence) is a local "Work Model" for AI coding agents. It sits underneath the agent you already use, such as Claude Code, Codex, OpenCode or Cursor, and keeps an accountable record of the agent's work on hardware you control.

AccInt tracks open promises (what the agent said it would do), evidence, approval gates and outcome credit, and turns successful runs into reusable runtimes, so the same kind of task gets better and cheaper each time it runs. When someone asks "why did the agent do that?", the reasoning and evidence are already written down. It needs no account and no API key, and everything stays on your machine.

Good fit if you: run AI coding agents on repeated work and want memory, accountability and approvals that you own.
Less ideal if you: only use AI chat occasionally; AccInt is built for ongoing agent workflows.

Pricing: free. It is an early release, so expect some rough edges.