Description
AI agents start every task from zero, repeating the same mistakes. autocontext is a recursive self-improving harness helping agents and their future iterations succeed on any task.
Give it a goal and it runs against evaluations, keeping lessons, dropping dead ends and leaving reports, playbooks, datasets and optional training artifacts.
Self-improving:Lessons kept.
Eval-driven:Repeated checks.
Playbooks:Knowledge saved.
Agents:Claude Code and Codex.
Give it a goal and it runs against evaluations, keeping lessons, dropping dead ends and leaving reports, playbooks, datasets and optional training artifacts.
Features
Self-improving:Lessons kept.
Eval-driven:Repeated checks.
Playbooks:Knowledge saved.
Agents:Claude Code and Codex.
