# AX Check: browser-use.com Checked 2026-09-18. Agents can install, connect, and price browser-use.com fast. 21 of 23 checks passed: quickstart, SDKs, API, MCP, and CLI all resolve, and pricing is posted openly with units and rates. ## Onboarding needs a login ## Coding sessions All three independent sessions (DeepSeek V4 Pro, Kimi K3, Qwen 3.8 Max) completed the task and found concrete pricing figures, including $0.02/browser-hour, $5/GB proxy, and per-model token rates, directly from the site's pricing and billing docs with stated assumptions. ### DeepSeek V4 Pro [View public run](https://agents.withgauge.com/p/runs/72262464-409b-4e36-a25d-86bd3bc2052a) · [Read transcript](https://www.ax-check.com/browser-use.com/sessions/deepseek.json) Final output gives concrete pricing ($0.02/browser-hour, $5/GB proxy, model token rates) sourced from browser-use.com/pricing.md (seq 41) with explicit assumptions called out (pay-as-you-go, free $15 signup credit, default model gpt-5.6-luna, BYOK fee). #### End-to-end onboarding - **Onboarding needs a login**: The agent researched Browser Use's docs, installed the SDK, read its source to confirm the v4 client API, and wrote a working example script. It never obtained an API key: there was none in the sandbox environment, and getting one requires an interactive OAuth signup (Google/GitHub/Microsoft) on the Browser Use Cloud dashboard, which the agent explicitly could not complete itself. No authenticated call to the real API was ever made — the script raised a ValueError for missing credentials and stopped there. Event 81: ```text BROWSER_USE_API_KEY set: False ``` Event 83: ```text The one hard block is **obtaining an API key** — there is none in this environment, and signup is an interactive OAuth flow I can't complete for you. ``` Event 97: ```text Set BROWSER_USE_API_KEY (get one at https://cloud.browser-use.com/settings?tab=api-keys&new=1). exit: 1 ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No API key available and signup requires human OAuth**: The sandbox had no BROWSER_USE_API_KEY, and creating one requires signing in via Google/GitHub/Microsoft on the Cloud dashboard — an interactive step the agent cannot self-serve. This is a normal credential/login limitation of the test environment, not a product defect: Browser Use documents the key requirement clearly and there is no way around it for a first-time developer other than signing up. Event 81: ```text BROWSER_USE_API_KEY set: False ``` Event 99: ```text I cannot actually execute the example: there's no `BROWSER_USE_API_KEY` in this environment, and getting one requires an interactive account signup (Google/GitHub/Microsoft OAuth at `cloud.browser-use.com`) — a human-required step I can't perform ``` #### Suggested Changes None identified in this transcript. ### Kimi K3 [View public run](https://agents.withgauge.com/p/runs/19c380e9-e8eb-4e20-8473-33e89b09c3e9) · [Read transcript](https://www.ax-check.com/browser-use.com/sessions/kimi.json) Pricing section cites concrete figures (credits, $0.24-0.36/1M tokens, 20% service fee, $0.02/hr browser, $5/GB network) sourced from browser-use.com/pricing (seq 22) and billing.md (seq 21), with pay-as-you-go/no-subscription assumption stated. #### End-to-end onboarding - **Onboarding needs a login**: Agent never obtained a real Browser Use API key. It probed the live API with a dummy key (got a legitimate 401), then verified its script only via a fully mocked run (fake urlopen response and fake SDK client), which is a local mock and does not count as an authenticated product operation. It explicitly stated it could not get a key itself since agent signup is disabled, and asked the human to supply one. Event 68: ```text HTTP 401 {"detail":"Invalid API key"} ``` Event 72: ```text with patch("urllib.request.urlopen", return_value=fake_resp), \ patch("browser_use_sdk.v4.BrowserUse", return_value=fake_client): ``` Event 74: ```text I can't run the example live — I have no API key, and I can't get one myself. ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No self-service API key; agent signup disabled**: Product requires a human-created API key from the Cloud dashboard. Agent signup endpoints are explicitly turned off, and the only agent-only alternative (x402 crypto payment) requires a wallet the agent does not have. This is a normal login/credential requirement, not a product defect, and it fully blocked live execution of the example. Event 31: ```text Agent signup is off. The `/cloud/signup` endpoints return `403`. ``` Event 74: ```text I can't run the example live — I have no API key, and I can't get one myself. ``` #### Suggested Changes - **Fix SDK version metadata so `browser_use_sdk.__version__` resolves**: Running `python3 -c "import browser_use_sdk; print(browser_use_sdk.__version__)"` raised AttributeError since the package has no `__version__` attribute. Add a `__version__` string to the top-level package `__init__.py` so basic sanity checks and version-logging code do not fail out of the box. Event 32: ```text AttributeError: module 'browser_use_sdk' has no attribute '__version__' ``` ### Qwen 3.8 Max [View public run](https://agents.withgauge.com/p/runs/a56171cb-57a9-416b-811b-8fcab414fb03) · [Read transcript](https://www.ax-check.com/browser-use.com/sessions/qwen.json) Final output's pricing section gives concrete figures (20% service fee, $0.02/hr browsers, $5/GB proxy, $5 min top-up) tied to named assumptions: BYOK vs managed billing, agent-vs-browser cost split, and concurrency tiers by lifetime spend. #### End-to-end onboarding - **Onboarding needs a login**: The agent never obtained a Browser Use API key. It checked the environment for one, found none, and explicitly stopped short of a live authenticated call. It only validated SDK imports, method existence, and script syntax locally with a dummy placeholder key string, which is not an authenticated product operation. Event 41: ```text no API key in env ``` Event 60: ```text Only one, and it's expected: **no API key in this environment**, so I couldn't execute a live run end-to-end. ``` Event 58: ```text syntax OK runs.create: True runs.wait_for_completion: True browsers.create: True browsers.stop: True ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No API key available to run a live example**: The sandbox environment had no BROWSER_USE_API_KEY, so the agent could not execute an authenticated call against the hosted Browser Use API. This is a normal credential-provisioning gap in the test environment, not a product defect — the agent correctly identified it and stopped rather than faking success. Event 41: ```text no API key in env ``` Event 60: ```text no API key in this environment, so I couldn't execute a live run end-to-end ``` #### Suggested Changes None identified in this transcript. ### Task given to each agent Help me build a simple example using Browser Use. Tell me how pricing works, and briefly tell me whether this product will be easy for you to manage. Let me know if you get blocked. If this product has no developer workflow you can act on, say so plainly and stop. Stay light: use the hosted product through its SDK or API. Do not start local service stacks or wait for long-running commands; if the quickstart requires either, say so plainly and stop. No product credentials were supplied and no purchases were authorized. ## Score: B · 84/100 (provisional) Grades come from completed site checks. Coding sessions and skipped checks do not affect the score. ### Clarity - **Failed** — Homepage answers Markdown requests ```text Homepage returned text/html even when text/markdown was requested; no Markdown representation served. ``` - **Pass** — llms.txt provides an actionable documentation index ```text llms.txt links docs, quickstarts, API, OpenAPI, SDK, MCP, pricing and skills with clear organization. ``` - **Pass** — llms.txt provides navigation guidance ```text llms.txt gives product-choice tables, developer URLs and starting guidance for agents. ``` - **Pass** — llms.txt mentions offered API, MCP, and skills ```text llms.txt mentions API V4, OpenAPI spec, MCP server, SDK and SKILL.md skills. ``` - **Pass** — A compact guide representation exists ```text SKILL.md and /skill serve a compact agent setup guide in Markdown. ``` - **Pass** — A focused guide is directly retrievable ```text https://browser-use.com/SKILL.md returns 200 text/markdown with full setup steps. ``` - **Pass** — Equivalent instructions fit a token budget ```text SKILL.md is 1572 tokens, well under the 8000-token budget. ``` - **Skipped** — Product-docs links survive format changes ```text Homepage Markdown unsupported, so link preservation across formats is unmeasured. ``` - **Pass** — The compact guide is independently actionable ```text SKILL.md gives ordered preflight, install, skill registration, browser connect, and end-to-end verification steps. ``` - **Pass** — Install and next-step links resolve ```text Fetched install/quickstart pages (SKILL.md, cloud quickstart, agent quickstart) all returned HTTP 200. ``` ### Onboarding - **Pass** — Docs lead to a relevant quickstart ```text Docs quickstart links to agent and browser quickstarts with concrete first steps. ``` - **Pass** — Installation commands are extractable ```text Install commands shown: pip install browser-use-sdk playwright, npm install browser-use-sdk@latest. ``` - **Pass** — Code examples are available without interaction ```text Python, TypeScript and curl code examples shown inline without interaction. ``` - **Pass** — Prerequisites and auth boundaries are explicit ```text API key creation link and X-Browser-Use-API-Key auth boundary stated explicitly. ``` ### Pricing - **Pass** — Pricing is readable without interaction ```text Pricing page renders as static Markdown with all plan tables and rates visible, no interaction needed. ``` - **Pass** — Prices are stated, not gated ```text Prices stated openly: $0.02/browser-hour, $5/GB proxy, per-model token rates, top-up amounts. ``` - **Pass** — Pricing units and limits are explicit ```text Units explicit: per browser-hour, per GB, per 1M tokens, concurrency tiers by lifetime spend. ``` - **Pass** — Agents identify pricing and its assumptions ```text 3 of 3 sessions were judged on pricing; 0 fell short. DeepSeek V4 Pro: Final output gives concrete pricing ($0.02/browser-hour, $5/GB proxy, model token rates) sourced from browser-use.com/pricing.md (seq 41) with explicit assumptions called out (pay-as-you-go, free $15 signup credit, default model gpt-5.6-luna, BYOK fee). Kimi K3: Pricing section cites concrete figures (credits, $0.24-0.36/1M tokens, 20% service fee, $0.02/hr browser, $5/GB network) sourced from browser-use.com/pricing (seq 22) and billing.md (seq 21), with pay-as-you-go/no-subscription assumption stated. Qwen 3.8 Max: Final output's pricing section gives concrete figures (20% service fee, $0.02/hr browsers, $5/GB proxy, $5 min top-up) tied to named assumptions: BYOK vs managed billing, agent-vs-browser cost split, and concurrency tiers by lifetime spend. This behavioural item does not affect the fast grade. ``` ### Activation - **Pass** — An API reference or OpenAPI spec is reachable ```text API V4 reference and OpenAPI 3.1 spec (33 paths) both fetched successfully. ``` - **Pass** — An MCP server is documented and well-formed ```text Hosted MCP server URL, auth header, client configs and tool list documented. ``` - **Pass** — A CLI install path is documented ```text CLI install documented: uv tool install browser-use, uvx browser-use, skill install. ``` - **Pass** — SDK packages resolve on their registries ```text PyPI browser-use, PyPI browser-use-sdk and npm browser-use-sdk all returned HTTP 200. ``` - **Pass** — Agent skills are published ```text SKILL.md and /skill publish an installable agent skill with setup steps. ``` [Full report data](https://www.ax-check.com/browser-use.com/report.json)