ClawBench

Open-source benchmark for browser AI agents on daily tasks.

Runs with mockssource: GitHubhomepagePythonApache-2.0commit 187cd252bc60

Python, Apache-2.0 licensed. The project labels itself: agent evaluation, agentic ai, ai agent benchmark, ai agents, benchmark, browser agent, browser automation and browser use.

ClawBench runs, with stand-ins for the services it depends on. An Argusic agent installed it in 7 minutes and hit 1 error on a clean machine with no GPU, and the whole session was recorded.

Python package clawbench-eval v0.10.0 installs and builds; all CLIs (clawbench, clawbench-run, clawbench-batch, clawbench-rescore, clawbench-analyze, clawbench-reproduce, clawbench-harbor-adapt, clawbench-edgebench-adapt, clawbench-edgebench-judge, clawbench-prorl-submit, clawbench-prorl-mock) are importable and respond to --help; batch dry-run produces correct 20-job matrix for v1-lite;

What the agent ended up with on a clean machine, in its own words. How this is measured

Measured by Argusic on a fresh machine every time. Every number links to its evidence. Argusic Score 76 of 100 (the mean of this project's run scores).

At a glance

verdict
Runs with mocks
Argusic Score
76 / 100
cost of the verifying run
$0.04 (measured)
recorded runs
2
last tested
stars
1,331
forks
79
open issues
136
watchers
14
size
5 MB
created
last push

Subject data from GitHub, linked at the top of this page, refreshed . Test data by Argusic (CC BY 4.0); every number links to a run page with the full log, the recording, and their sha256 hashes.

What was measured

From the run that decided this verdict. Every line is a recorded value, not an estimate.

Time lost to each failure while testing ClawBench, drawn from the recorded run
Every bar is a failure the agent recorded, with the time it cost. Drawn from this run, not from anywhere else.

What went wrong, in the agent's own words

The agent's own words, worst first, never rewritten into advice.

  • One test (test_batch_stays_importable_without_a_container_engine) shows non-deterministic ordering sensitivity — fails when run in isolation after other host_timeout tests but passes in full suite run. Not a code defect, test isolation edge case.1 minute
Install time
7 minutes
Cold machine to finish
3 minutes
Errors hit and fixed
1 hit, 0 fixed with no human help
How the result was proved
uv run pytest -q --tb=short => 320 passed, 2 skipped
Model tokens used
47,608
Exact commit tested
187cd252bc60
Container image
sha256:cdd920bce783

What was not measured

The other half, so you can see exactly how far this evidence goes. A gap is printed as a gap and never filled in.

Peak GPU memory
this ran on a CPU machine, so there was no GPU to sample
Behaviour over a long run
each attempt uses the smallest documented settings and a fixed time budget, so this measures whether it runs, not how it holds up under load
Behaviour against real services
this run was verified against stand-in services only

The end of the recording

The last lines of the terminal session that decided this verdict, exactly as they were recorded.

tokens used
47,608
Here's my final assessment:
**Setup:** Installed `uv`, ran `uv sync` to set up the venv and install all deps, copied `models/mo…
**Verification:**
- Full test suite: **320 passed, 2 skipped** (skips are expected — `harbor` and `evalport-sdk` are …
- All 301 task.json files (v1: 152, v2: 129, v1-lite: 20) are valid JSON
- `clawbench-batch --dry-run` correctly computes the 20-job matrix for v1-lite
- All 11 CLI entrypoints import and show help
- The TUI and all runner modules import without errors
- Model config loads both example models from `models.yaml`
**Note on container dependency:** Docker/Podman are not installed in the container, so real contain…

Replay the whole session, every command from a clean machine to this point.

How it was tested

One agent, one container, no human help. It installs the project the way its own documentation says to, runs it, and fixes what breaks. Everything below is recorded as it happened: the terminal session, the log and the exact commit. The full method.

Strengths and limits

Measured facts, not opinions. How this is written.

What went well

  • Installed in 7 minutes, faster than the median of the 4 comparable projects Argusic has measured.
  • Licensed Apache-2.0, as reported by its host.
  • Measured 2 times, so the result is not a one-off.

What did not, or is not known

  • Only reached a running state against stand-in services, so nothing here proves it works against the real ones.
  • Hit 1 error during setup, 1 of which were not solved.

Of the 5 agent-evaluation projects Argusic has installed and timed, ClawBench was the 3rd fastest to reach a running state, and 3 of 5 reached one at all.

Also tested, in the same area

Every one of these was installed and run by Argusic on a clean machine. Nothing appears here that was not tested.

Run history

tested byattemptoutcomescorecost USDwhen
Argusic Runner2/3Unverified80.000.02
Argusic Runner1/3Runs with mocks72.000.04

Topics (from GitHub)

agent-evaluationagentic-aiai-agent-benchmarkai-agentsbenchmarkbrowser-agentbrowser-automationbrowser-usechrome-agentchrome-extensioncomputer-usedatasetevaluationeveryday-tasksllmllm-evaluationonline-tasksreal-world-benchmarkweb-agentweb-agents

Embed the badge

Markdown for the project README. It links back here; terms on the terms page.

[![Tested by Argusic](https://argusic.com/badge/ClawBench.svg)](https://argusic.com/subject/clawbench)

Questions

Does ClawBench run?
ClawBench runs, with mocks standing in for external services. Argusic installed and launched it on a clean machine in 7 minutes, hitting 1 error on the way, and recorded the session.
How did Argusic test ClawBench?
On a fresh, disposable machine, with every command recorded and the repository pinned at commit 187cd252bc60. 2 attempts are recorded, and the full method is on the methodology page.
What did it cost to test ClawBench?
The run that produced this verdict cost $0.04: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
How long does ClawBench take to install?
7 minutes in the run that decided this verdict, from a clean machine to a running state. That is one measured attempt, not an average.
Does ClawBench need a GPU?
Not to start. Argusic reached a running state on a machine with no GPU.
What goes wrong when installing ClawBench?
1 thing broke in the recorded run. Each one, and the time it cost, is listed on this page.
How does ClawBench compare with the alternatives?
Of the 5 agent-evaluation projects Argusic has installed and timed, ClawBench was the 3rd fastest to reach a running state, and 3 of 5 reached one at all.
Where is the evidence for ClawBench?
All 2 recorded runs are on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.

Discussion