Claude Autoresearch Mod — Development skill for Claude Code
Autonomous experiment loops for Claude Code ported from pi-autoresearch.
How to install Claude Autoresearch Mod
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open bn-l/claude-autoresearch-mod and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Autoresearch Mod does
Autonomous experiment loops for Claude Code ported from pi-autoresearch.
Alternatives in Development
- Om Tidy — Self-maintenance pass — acts on every hygiene flag: archives completed work, groups loose clusters, splits ove 4.6k ★
- Autoresearch — Claude Autoresearch Skill — Autonomous goal-directed iteration for Claude Code 2.5k ★
- Evo — turns your codebase into an autoresearch loop — discovers what to measure, instruments the benchmark, then run 1.4k ★
README
pi-autoresearch for Claude Code
Uses the new mods api to change claude code's UI for the same autoresearch experience as in pi.
A Claude Code mod that runs autonomous optimization loops. It will try an idea, benchmark it, keep improvements, revert regressions and repeat this process in a structured way.
For more information see, [pi-autoresearch](https://github.com/davebcn87/pi-autoresearch), this is an almost 1:1 port of that (removing some small bugs).
Inspired by [karpathy/autoresearch](https://github.com/karpathy/autoresearch).
The session format is the same as pi-autoresearch (`.auto/log.jsonl`) and can be used interchangedly.
Quick start
claude plugin install autoresearch --marketplace bn-l/claude-autoresearch-mod
To update: `claude plugin update autoresearch@bn-l`, or turn on auto-update for bn-l in `/plugin` → Marketplaces.
To try it on a toy project first, copy [`examples/sort-bench`](examples/sort-bench) somewhere, make it a git repository, and run `/autoresearch make sort.js faster` in it.
Requirements
Claude Code **2.1.292** or newer.
The browser dashboard (very optional) requires `node` (18 or newer) or `bun`.
Options
| Option | Default | What it does |
|---|---|---|
compactAtPercent |
70 | Compacts between iterations once the context is this full (0 turns it off) |
questionWaitMinutes |
5 | When a turn ends by asking you something, waits this long for your reply before carrying on (0 doesn't wait) |
`claude plugin configure autoresearch` shows them. To set one when installing, add `--config questionWaitMinutes=0` to the install command; to change one later:
echo '{"questionWaitMinutes": "0"}' | claude plugin configure autoresearch --values-stdin
Related Skills
Run Autoresearch
Run an autonomous experiment loop via the Autoresearch Orchestrator agent
Vault Autoresearch
3-round autonomous research loop with intermediate notes and synthesis
Agent Usage Guard
A Claude Code mod that holds subagent fan-out, heavy-context prompts and retry loops before they spend your us
Animus Sprint Loops
A lightweight, platform agnostic protocol for autonomous development and objective handling (repo acts as a Cl
Feature Implementation
Autonomous feature implementation with verification loops.
Ldd Autonomy Ack
Pre-grant, list, check, or revoke autonomy acks so architect-mode's creativity=inventive can fire in autonomou
Related Agents
Autoresearch Orchestrator
Use this agent to run an autoresearch experiment session end-to-end — setup, baseline, and a batch of edit→mea
Loop Operator
Safety net for autonomous agent loops. Detects stalls, reduces scope on repeated failures, escalates to user w
Sweep Controller
Controls autonomous fix-and-revalidate loops until all journeys pass or max attempts exhausted