# evo-hq/evo

turns your codebase into an autoresearch loop — discovers what to measure, instruments the benchmark, then runs tree search with parallel subagents.

Repository: https://github.com/evo-hq/evo
Canonical: https://ross.abutalabs.com/products/evo-hq-evo
Homepage: https://evo-hq.com
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agent-skills, autonomous-agents, autoresearch, claude-code, code-optimization, codex, evolutionary-algorithms, llm-agents
Last push: 2026-07-17T16:52:59+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 93, release rhythm 93, longevity 10
- inputs: {"age_days": 150, "days_push": 47, "days_rel": 47, "gap_med": 2.5, "n_releases_24m": 19}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1421, forks 109 (observed 2026-08-28T04:04:40.679037+00:00)

## What it is
Evo is a plugin for agentic coding frameworks (Claude Code, Codex, Cursor, etc.) that turns a codebase into an autoresearch loop: it discovers what to measure, instruments benchmarks, and runs tree search with parallel subagents in isolated git worktrees. Experiments that improve a metric are kept, regressions are discarded, and gating plus a dashboard provide safety and observability.

## Use cases
- automatically optimize my codebase with an LLM agent loop
- run parallel AI agents to improve a benchmark score
- discover what metrics to optimize in my repo
- set up an autoresearch hill climb with tree search
- optimize a JSON parser speed automatically
- run experiments in git worktrees with subagents
- add regression gates to autonomous code experiments
- monitor autonomous optimization experiments on a dashboard

## When to choose
- you use Claude Code, Codex, Cursor, or a similar agentic CLI and want autonomous code optimization
- you have a measurable benchmark or metric and want an LLM to iteratively improve it
- you want parallel exploration via tree search instead of a single greedy hill climb
- you need regression gates and observability around autonomous experiments

## When to avoid
- you have no objective metric or benchmark to optimize against
- your codebase cannot run safely in automated experiments or sandboxes
- you need a hosted inference-routing product rather than a local optimization loop (the evo-hq.com Router is a separate offering)
- you want fully deterministic, non-LLM optimization

## Facets
- artifact type: plugin
- maturity: active
- function: agent-framework, benchmarking, developer-tools, cli, workflow-automation, llm-inference
- domain: developer-tools, artificial-intelligence, performance
- platform: python, cli, cross-platform, cloud
- tags: autoresearch, evolutionary-algorithms, tree-search, claude-code, codex, code-optimization, subagents, git-worktrees, llm-agents, experiment-loop, automation, ai-agents, docker

## Member repositories
- evo-hq/evo (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:40.679037+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 2026-08-30T04:37:45.093898+00:00, confidence not recorded.
  - readme: https://github.com/evo-hq/evo (fetched 2026-08-28T04:04:40.679037+00:00, sha 9ec35193c090)
  - homepage: https://evo-hq.com (fetched 2026-08-29T11:50:15.984585+00:00, sha 66f2287281d6)
  - site_page: https://evo-hq.com/docs (fetched 2026-08-29T11:50:16.030504+00:00, sha a84993152190)
- Data as of 2026-08-30T08:39:29.467469+00:00.