ai-agent

ClawProBench

Name: ClawProBench
Author: suyoumo

ClawProBench is a live-first benchmark harness for evaluating LLM agents in the OpenClaw runtime with deterministic grading and repeated-trial reliability.

823 stars54 forksRustUpdated 8/25/2026100% free · open source

What it does

ClawProBench runs repeatable, deterministic benchmarks on LLM agents inside the OpenClaw runtime so you can measure performance with confidence.

When to use it

•You need to compare multiple LLM‑agent implementations before picking one for production.
•Your product relies on OpenClaw‑compatible agents and you want reliable, repeatable grading across runs.
•You want to embed a benchmark suite into CI/CD to catch regressions in agent behavior.

Ready-to-paste prompt

clawprobench run --bench examples/bench.yaml --agent ./agents/customer‑support-agent --repeat 5 --output results/cs_report.json

Heads up: ClawProBench requires the OpenClaw runtime (v0.7 or later) to be installed and available on the PATH; without it the `run` command will exit with “OpenClaw not found”.

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🛠️ Technical setup

Expect 20–40 minutes in a terminal — or let your AI agent drive it.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/clawprobench && curl -fsSL https://workflowstacks.com/api/skills/clawprobench/claude-skill -o ~/.claude/skills/clawprobench/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How ClawProBench works

Codeflow

Free to inspect

ClawProBench is a very large Rust project (~1.2M lines across 2782 code files, plus 773 test files). It is a full software project: use it through its install path rather than reading it end to end. Last commit a month ago, Apache-2.0 license, has a test suite.

Size

Very large codebase

~1.2M lines · 2782 code files · days to read — use, don't read

Setup

Developer setup

A real software project. Use it via its install path; don't expect to read it all.

Runs on

Python

No API keys detected

Rust 80%Python 8%TypeScript 8%JavaScript 1%Shell 1%

Where to start reading

1
README.md
Start here — what it does and how to install it
2
run.py
Where the program starts running
3
requirements.txt
Dependencies and the commands it exposes
4
ironclaw/src/app.rs
Inside ironclaw/src/ — the main logic begins here

What's in each folder

ironclaw/src/Core code — the actual logic396 files

ironclaw/skills/Prompts, skills & agent definitions40 files

ironclaw/prompts/Prompts, skills & agent definitions2 files

ironclaw/docs/Documentation407 files

docs/Documentation8 files

ironclaw/crates/Folder2502 files

nanoclaw/Folder489 files

datasets/Data files461 files

READMEHas testsDocumentedCI checksApache-2.0 licenseUpdated this month

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🛠️ Technical setup

Expect 20–40 minutes in a terminal — or let your AI agent drive it.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/clawprobench && curl -fsSL https://workflowstacks.com/api/skills/clawprobench/claude-skill -o ~/.claude/skills/clawprobench/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

Quick Actions

Details

Creator

suyoumo

Language

Rust