Awesome Computer-Use Agents

A source-backed collection of papers, models, datasets, benchmarks and tools for agents that use computers.

English · 简体中文

Search & filter · Research papers · Browse topics · Data & agents

Read the research survey — Methods, evaluation, human agency, and reusable research assets. English working manuscript with a versioned bibliography and reading register.

Find by task · Explore · Highlights · Recently reviewed · Use the data

Coverage ledger · Contribute · GitHub organization

Discovery is scheduled three times daily at 00:17, 08:17 and 16:17 UTC. Actual scan results · Schedule and safeguards.

For people, RAG systems, and agents — clone the complete catalog:

git clone https://github.com/reacher-z/Awesome-CUA.git

Agent instructions · RAG and agent guide · Complete JSONL · Retrieval documents

The goal is comprehensive coverage of public research and industry resources, with provenance that can be inspected. Papers, models, training data, evaluation environments, frameworks, infrastructure, products, and publishers belong in the same searchable catalog.

Global completeness is a goal, not a verified claim. We publish the source populations, review states, and remaining gaps. Bibliographic metadata verification is different from reviewing a resource's relevance or quality.

Browse by topic · Topic subsets and fingerprints

cd Awesome-CUA
python3 scripts/catalog.py search "grounding" --kind paper --status curated --limit 0 --json
python3 scripts/catalog.py search --kind benchmark --environment web --status curated --limit 0 --json
python3 scripts/catalog.py search --kind dataset --environment android --status curated --limit 0 --json
python3 scripts/catalog.py search --kind framework --status curated --limit 0 --json
python3 scripts/export_rag.py --output dist/rag-documents.jsonl
python3 scripts/catalog.py get arxiv:2404.07972 --related --json

3,260 resource records · 1,671 curated · 2,789 paper and survey records · 54 discovery documents

Evidence through 2026-10-24T04:38:00Z. 1,589 records have bibliographic verification only; 907 discovered identifiers remain unresolved, not counting 136 links reviewed as non-resources or other addresses of indexed resources. 138 scope exclusions retain their original metadata in the archive. Counts overlap as explained below.

By platform: Web 646 · Desktop 334 · Windows 65 · macOS 30 · Linux 54 · Android 323 · iOS 22 · HarmonyOS 3 · Cross-platform 197

Coverage spans web, desktop, Android, iOS, Windows, macOS, Linux, and cross-platform work. Topics include grounding, screen understanding, planning, memory, long-horizon workflows, reinforcement learning, data generation, safety, robustness, and evaluation. Environment facets overlap; framework support is not assumed from a model's benchmark results.

The most-mentioned curated resources per category: the count is how many other cataloged resources mention each one in their reviewed first-party source (README, model or dataset card). It reflects connectedness inside this catalog, not quality, popularity, or benchmark performance. Every category page lists all reviewed entries. Surveys come first, newest first.

Start with the newest surveys · all 19 →

Benchmarks · all 89 →

Models · all 98 →

Datasets · all 86 →

Frameworks · all 79 →

Agents · all 45 →

Newest editorial approvals of models, datasets, benchmarks, systems, and tools (newest first). New papers appear by date in the topic hubs.

Browse the interactive website to search and filter all resources, inspect source evidence, switch between English and Chinese, or download the complete catalog. Website development and publishing.

The English data hub and 中文数据中心 organize public downloads into six groups: complete catalog, RAG documents, citations, topics and comparisons, sources and review status, and schemas. Cloning still gives you the complete source catalog; the hub provides exports from its published snapshot.

The download manifest records each listed export's actual count and unit, scope, byte size, and SHA-256 fingerprint. These are different formats and overlapping views of the catalog, so their counts must not be added together. Compare catalog_sha256 across manifests before combining files from different downloads.

Reviewed research · Computer-use agent guides · Static resource directory · Search and AI discovery

• Catalog JSONL: one resource per line, with stable IDs, typed fields, provenance, scoped licenses, and review status.
• Reference graph: outbound references from primary sources. mentions does not mean trained_on or evaluated_on.
• Discovery registry and candidates: source snapshots and where each identifier was found.
• Review queue: unresolved evidence, unindexed discoveries, and proposals requiring editorial work. Discovered links reviewed as non-resources or as other addresses of indexed resources stay listed with a resolution and their reason.
• JSON Schema and coverage JSON: machine-checkable contracts and coverage accounting.
• BibTeX and CSL JSON: reviewed-paper citations with recorded authors, dates, and versions. Each resource page also provides its own record.json.

# Python 3.11+. Search and export require only the standard library.
cd Awesome-CUA
python3 scripts/catalog.py search "grounding" --kind model
python3 scripts/catalog.py search --kind dataset --environment android
python3 scripts/catalog.py search --kind product --status curated --json
python3 scripts/catalog.py get arxiv:2404.07972 --related --json
python3 scripts/catalog.py build
# Creates dist/catalog.sqlite and dist/catalog.json, with no network calls.
python3 scripts/export_rag.py --output dist/rag-documents.jsonl
python3 scripts/export_citations.py --format bibtex --output dist/citations.bib
# Add --include-metadata to include every active record, with its review state intact.

Use SQLite to join resources, environment/topic facets, and reviewed relations. See query examples. data/catalog.jsonl is committed; SQLite and JSON array exports are reproducible local build products.

Counts refer to resources, not unique projects. A paper, repository, dataset release, model checkpoint, product, and publisher can be separate related resources. Model families and individual checkpoints are explicitly linked where reviewed.

Code, data, weights, and document licenses are separate. Missing values are null. Dataset archive counts are not converted into training-example counts. Scores from different benchmark versions, splits, action budgets, or evaluators are not combined into a ranking.

python3 -m venv .venv
.venv/bin/python -m pip install --require-hashes -r requirements-dev.txt
.venv/bin/python scripts/catalog.py validate
.venv/bin/python scripts/catalog.py build --check
.venv/bin/python -m unittest discover -s tests -v
.venv/bin/python scripts/catalog.py build
node --test tests/web.test.mjs
.venv/bin/python scripts/build_site.py

These commands match the checks that Catalog quality runs on every push and pull request, in the same order. They need Python 3.11+ and Node.js 24+, the engines version in package.json (CI uses Python 3.12 and Node.js 24). The web tests use only the built-in Node.js test runner, so nothing needs to be installed with npm. npm run test:web and npm run build:web are shortcuts for the last two commands.

Network discovery and verification are explicit commands; ordinary builds are offline. Changed primary snapshots require another editorial review before publication. Maintenance guide · Methodology · Data dictionary

The source registry credits the community collections and surveys used for discovery. Bibliographic records use retained arXiv, publisher, and author sources; publication observations preserve edition-specific titles, authors, and date precision. Selected assets are checked against publisher repositories, documentation, and model/dataset cards. Upstream READMEs, papers, model weights, and datasets are not redistributed here.

Original repository code and annotations use the MIT license. Linked resources retain their own terms; this repository does not relicense them.

To cite the catalog, use CITATION.cff or GitHub's "Cite this repository" button. The data changes with every scan, so pin the exact version you used. Each v* tag has a data release with versioned downloads built from the tagged commit: the catalog as JSONL, JSON and SQLite, the coverage ledger, topic and profile exports, and their SHA256SUMS. Cite the release tag, or record the commit from git rev-parse HEAD in your clone, or link files at that commit, such as https://github.com/reacher-z/Awesome-CUA/blob/<commit>/data/catalog.jsonl. Cutting a release.