# greyhaven-ai/autocontext

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

Repository: https://github.com/greyhaven-ai/autocontext
Canonical: https://ross.abutalabs.com/products/autocontext
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agents, ai, autoresearch, llms, ml, openclaw, claude, claude-code, codex, hermes, hermes-agent, pi, pi-coding-agent
Last push: 2026-08-26T18:22:53+00:00

## Health v2 (maintenance only)
Score: 60/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 35, longevity 14
- inputs: {"age_days": 203, "days_push": 7, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1286, forks 109 (observed 2026-08-28T04:04:14.943891+00:00)

## What it is
Autocontext is a recursive self-improving harness that runs AI agents against evaluations, retains useful lessons, and produces traces, reports, playbooks, datasets, and optional local-model training artifacts for subsequent runs. It ships as a Python CLI (autoctx) with a TypeScript/Node CLI variant and a Pi editor extension.

## Use cases
- iteratively improve agent performance on a task across runs
- run an agent against an evaluation and keep the lessons learned
- generate playbooks and datasets from agent runs for future iterations
- optimize prompts and strategies automatically over multiple iterations
- train local models from accumulated agent run artifacts
- improve customer-support reply quality with an automated agent loop

## When to choose
- you want an automated evaluate-learn-retry loop for LLM agents
- you use Claude Code, Codex, Pi, or OpenAI-compatible agents and want cross-run improvement
- you want traces, reports, and datasets persisted between agent iterations

## When to avoid
- you need a simple one-shot agent runner without iterative optimization
- you want a GUI-driven agent builder rather than a CLI harness
- your workflow cannot share state or artifacts between runs

## Facets
- artifact type: cli-tool
- maturity: active
- function: agent-framework, llm-training, benchmarking, cli, workflow-automation
- domain: large-language-models, artificial-intelligence, developer-tools
- platform: python, cli, cross-platform
- tags: self-improving-agents, agent-harness, evaluation-loop, iterative-optimization, claude-code, codex, agent-traces, playbooks, ai-agents, automation, nodejs

## Member repositories
- greyhaven-ai/autocontext (main) score 60

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:14.943891+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:56:34.965141+00:00, confidence not recorded.
  - readme: https://github.com/greyhaven-ai/autocontext (fetched 2026-08-28T04:04:14.943891+00:00, sha 92648b0c96e4)
  - registry_pypi: https://pypi.org/pypi/autocontext/json (fetched 2026-08-29T12:12:13.992970+00:00, sha ab667b982599)
- Data as of 2026-08-30T08:39:29.467469+00:00.