Describe a goal. Test the real app.
Describe what to do and what should happen. Let AI operate the real app and verify the outcome. Inspect screenshots, actions, and results to see what happened.
An open-source project from FSQ Lab.
Website · Quick start · Benchmarks · 中文 · Guide · Contribute
See a complete test
https://github.com/user-attachments/assets/3acfecf4-bf48-4c76-8245-2eeef0dedb1f
YouTube · English captions · Media provenance
A real Android Quick Test, from goal to evidence. Try Web below; setup and execution differ across platforms, and search results and outcomes can vary. This recording is an observation, not a success guarantee.
Benchmarks
95.69% (111/116, pass@1) on AndroidWorld · Single Model · Strict Budget · 0 App Skills.
Representative open-source and public-code snapshots from the November 2026 Google Research archive and public reports, plus a human reference—not the complete leaderboard including closed-source submissions or an official same-protocol ranking. Protocols, step budgets, and score-to-code commit bindings differ. Phone Use's license is unverified; UI-Venus/UI-TARS provide weights and inference code, not a verified complete score-producing AndroidWorld Harness. License labels do not certify the full evaluated stack; using a commercial model API alone does not make an Agent closed source. Minitap's 94.0% is pass³, not pass@1.
FSQ uses 113 original tasks + three disclosed task-contract corrections and mode B, not unmodified-upstream
or stock-strict grading. The image rounds 95.69% to 95.7%; 111/116* refers to these conditions.
Full report · METHOD: frozen versions, budgets and audit limits · Comparison provenance
Quick start
Choose Coding Agent or FSQ UI—equal paths to one Quick Test and evidence review. You need Python 3.11+, a supported configured Provider, an explicitly selected model, and platform prerequisites for Web, Android, Windows, or macOS. Enter credentials privately in local FSQ Settings, never in Coding Agent chat.
- Coding Agent: follow the Skill setup guide, then reload your agent. Ask it to use FSQ Quick Test with Web, your selected available browser, and the complete goal below.
- FSQ UI: install and launch below, configure Provider/model in Settings, then choose Quick Test, select Web and an available browser, paste the goal, and select Start Quick Test. UI guide.
python -m pip install --upgrade fsq-agent
fsq ui
Use the selected browser to search for the FSQ-lab/FSQ GitHub repository. Open that repository, follow the website link in its About section, and verify that the destination page displays both “Describe a goal.” and “Test the real app.” Report the outcome with the captured page evidence. Only browse public pages. Do not sign in, change repository content, or submit forms. Stop and report a blocker if the repository or About website link cannot be identified.
Quick Test prepares ~/.fsq/quick-test/ without manual project Workspace setup; it neither saves a formal Case
nor runs Strict Replay. Open the report and inspect captured evidence: failed, inconclusive, or preparation-error
outcomes do not mean the checks passed.
From goal to evidence
Describe the journey and checks. This editorial task card explains the recorded goal; its annotations are not product UI.
Watch the real app. The Android browser and FSQ action timeline connect actions with screenshots and interface state.
Inspect the result. Real page and same-Run report captures separate execution, verification, and evidence completeness; surrounding explanations are editorial, not product UI.
Go further when you need to
- Inspect: compare screenshots, interface state, events, and reports with your expectations.
- Manage Cases: FSQ Case Manager supports registered Workspaces, explicit Case creation or saving an available candidate, and run history. Case generation is conditional; not every run produces one.
- Replay separately: review any returned Case unchanged, then explicitly request Strict Replay. It operates the target again and requires prerequisite and side-effect checks; Quick Test never runs it automatically.
Help shape FSQ
If useful, Star FSQ or share your experience—neither is required to try it. Include version, platform, sanitized goal, expected/actual outcome, and reviewed diagnostic excerpts only. Remove credentials, personal data, private URLs, paths, and device identifiers from screenshots/logs; use Security for security issues.
Alpha: for evaluation and contribution; interfaces may evolve before 1.0. Evidence stays local and may contain private app data: review before sharing, and never commit evidence or credentials. AI operations send necessary provider-bound inputs to your configured Provider. See support and stability.
Documentation
CLI reference · Case format · Advanced Codex compatibility
License
MIT — Copyright (c) Microsoft Corporation.
Metadata
Release files for fsq-agent 0.3.5
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| fsq_agent-0.3.5.tar.gz | 12.4 MB | Without provenance | Details for fsq_agent-0.3.5.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| fsq_agent-0.3.5-py3-none-any.whl | Python 3 | none | any | Details for fsq_agent-0.3.5-py3-none-any.whl |
Total release size: 13.6 MB
Release files / fsq_agent-0.3.5.tar.gz
| Download URL | fsq_agent-0.3.5.tar.gz |
|---|---|
| Size | 12.4 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
1268e296c576530a2325bb0375ca1002a77e05539d88ccac3a480a309f4afb5e
|
|
BLAKE2b-256 checksum How to use checksums |
9eb9f95eb31954aca38d3b2f7eddaee68e9e86dac6687802db16b4fe0179b361
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / fsq_agent-0.3.5-py3-none-any.whl
| Download URL | fsq_agent-0.3.5-py3-none-any.whl |
|---|---|
| Size | 1.2 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
e47874ad957360142d3cd4e5ace6269a247fe857c9d89cf961136b8505ece9ed
|
|
BLAKE2b-256 checksum How to use checksums |
0617a5fb5d1bdfaf957c40e41b58f2967f54c10ee49731b7e4794d1e09313703
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for fsq-agent 0.3.5
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.