BigHugger
GH Repository · greyhaven-ai

autocontext

a recursive self-improving harness designed to help your agents (and future iterations of those agents) succeed on any task

stars
1,295
30-day movement
+10/day
Related entries
63
Connections
3
otherPythonagentsclaudeclaude-codeautoresearchmlaiopenclawhermespillmscodexhermes-agentpi-coding-agent

autocontext is a Python repository from greyhaven-ai described as a recursive self-improving harness for helping agents, and future iterations of those agents, succeed on tasks. It ships AGENTS.md and CLAUDE.md files, suggesting it provides context/configuration for agent tools like Claude Code and other coding agents listed in its topics.

Reach for it if you want a harness that iteratively improves the context your agents work from across runs.

Use it to

  • Set up AGENTS.md and CLAUDE.md context files for coding agents
  • Run agents across iterative self-improvement loops
  • Apply the harness to varied task types
  • Adapt it to agents like Claude Code, Codex, or Hermes

For Developers building or running LLM agents

Role
other
Language
Python
Licence
Apache-2.0
Forks
110
Open issues
1
Last push
2026-09-16
Skills shipped
3
topicsagentsllmsself-improvingclaude-codeautoresearchpython