Back to Ecosystem
Research & BenchmarksVerified

Claw-SWE-Bench

A benchmark for comparing heterogeneous agent harnesses—described by the authors as claws—on coding tasks under controlled settings.

Decision information

Verified: 8/12/2026
Ecosystem category
Research & Benchmarks
Provider
Research
Content type
Research
Pricing
Open Source
Open source
Yes
License
Not recorded
Maintenance
Active
Install methods
Not recorded
Platforms
Not recorded

Verification source: https://arxiv.org/abs/2606.12344

Why it matters\n\nClaw-SWE-Bench is evidence that “claw” is beginning to function as a category for agent harnesses beyond a single product. It is a benchmark, not a Claw agent.

Tags

researchbenchmarkcoding-agentsagent-harnessclawsswe-bench
Claw-SWE-Bench | All Claw