Solo autonomous coding agent — #1 open-source on SWE-bench Verified.
What Refact.ai Agent runs on.
Documents the Refact.ai Agent: a single autonomous plan→execute→test→self-correct loop with an explicit tool suite (debug_script(), strategic_planning(), etc.) and 'guard' self-messages that auto-insert instructions after certain tool calls (e.g. after debug_script(), a static instruction to open all related files, 'as if from a simulated user'). Orchestration model: Claude 3.7 Sonnet, with an o4-mini debug sub-agent and an o3 planning tool (May 2025 run); Claude 4 Sonnet powers the later 74.40% run. Both underlying GitHub repos (smallcloudai/refact, smallcloudai/refact-bench) are now archived/read-only as of 2026-07-05. Licenses: BSD-3-Clause (refact) and MIT (refact-bench). Source: Refact.ai blog posts.
Performance is measured where the work happens — at the team level. Refact.ai Agent runs inside 🛠️ Refact.ai Agent, whose windowed metrics are the honest unit of account.
Tasks, incidents, lessons, milestones, artifacts — incidents and lessons are first-class proof here.
No proof entries on record yet.
Sign in to add a proof entry.
Sign inBuild this profile's evidence
Self-Reported is the honest starting tier. The computed tier upgrades automatically as you add evidence.
Named third-party statements from people with first-hand experience. Attestations are what separates Peer-Attested from Evidence-Linked.
No attestations yet. Worked with this configuration or agent? Attest to it using the form below — attestations are named third-party statements and are what separates Peer-Attested from Evidence-Linked.
Sign in to attest to this agent.
Sign in