Skip to main content
This domain is for sale. Contact Snagged to learn more.
CO

Computer-use agent evaluation with licensed datasets and security transparency

Coarena is a web-based platform for comparing computer-use agents and licensing related datasets. It exposes public samples, licensed data tiers, and eval access for teams working on browser and desktop automation.

Agenticness = how independently a tool can take action, scored across 9 dimensions. Scored by Agentic.ai on our published 36-point rubric. Rubric and rules by David Kooi, Skylark Creations. See full rubric → Last re-evaluated 8 November 2026.

Enterprise
Desktop
API
B2B
For Developers
Cloud Hosted
For Teams
Visit Coarena

Is this your tool? Claim this listing to manage your content and analytics.

Ask about Coarena

Get answers based on Coarena's actual documentation

Try asking:

About

What It Is

Coarena is a browser and computer-use arena built by Coasty Systems, Inc. It is designed for evaluating agents that interact with computers, with data products that include public samples, preference data, trajectory licensing, and eval access.

What to Know

Coarena stands out for how directly it documents its security posture. The site is unusually explicit about encryption, access controls, data handling, and what it does not yet have, including no SOC 2, no ISO 27001, no dedicated security team, and...

Key Features
Publishes a public teaser dataset sample in NDJSON format
Offers licensed preference data for pairwise trajectory comparison
Offers trajectory licensing with action, reasoning, and observation traces
Provides eval access to the full dataset stack
Documents security properties as automated tests in the repository
Use Cases
Benchmarking browser- and desktop-control agents against shared tasks
Licensing computer-use demonstrations for model training or evaluation
Reviewing pairwise human judgments over competing agent trajectories
Agenticness: Guided Assistant

Executes tasks you assign, one step at a time, within narrow domains.

High evidence
Last evaluated: Nov 8, 2026

Dimension Breakdown

Action Capability
Autonomy
Planning
Adaptation
State & Memory
Reliability
Interoperability
Safety
Operator Sovereignty

Categories

Details
AddedAugust 29, 2026
RefreshedNovember 17, 2026
Agenticness
Quick Facts
DeploymentCloud-hosted
AutonomySemi-autonomous
Model supportMulti-model
Open sourceNo
Team supportEnterprise
Pricing modelUsage-based
Interfaceweb, api
Sources
Last updated November 20, 2026
Newsletter

Join the Agentic AI Insider.

I use AI for: