BrowserOps MCP
Browser QA
With Receipts
BrowserOps is a browser evidence and UI verification MCP for agents. This page is a reliability note, not a broad browser-automation benchmark. It tracks whether BrowserOps can observe pages, run explicit verification scenarios, collect browser diagnostics, and emit receipts reviewers can audit.
Current Reliability Gate
Checks Passed
7/7
Tool Surface
2 MCP
Receipt Types
2
Unsafe Targets
Blocked
Reliability Notes
Verification Checks
| Check | Status | Evidence |
|---|---|---|
| Deterministic observation | Passed | Fixture observation receipt with title, URL, text, links, DOM hash, screenshot hash, and evidence fingerprint. |
| Deterministic verification | Passed | Scenario verification receipt with explicit steps, assertions, console messages, network failures, and screenshot hash. |
| MCP stdio call | Passed | JSON-RPC initialize, tools/list, and tools/call path exercises browser_verify_url through the packaged server. |
| Receipt validation | Passed | browserops-validate-receipt validates both observation and verification receipt fingerprints without rerunning a browser. |
| Real Chromium observation | Passed | Optional Playwright smoke passes when BROWSEROPS_CHROMIUM_EXECUTABLE_PATH points at a working Chromium binary. |
| Real Chromium verification | Passed | Optional Playwright verification smoke fills a field, clicks a button, waits for text, and validates assertions. |
| Unsafe target refusal | Passed | Localhost, RFC1918, link-local metadata, .local hostnames, and browser-internal schemes are blocked by default. |
Product Surface
MCP tools
browser_observe_url, browser_verify_url
CLI commands
browserops-observe, browserops-verify, browserops-validate-receipt, browserops-doctor
Scenario input
Reusable JSON steps and assertions
Evidence artifacts
JSON receipts and screenshots
Browser backend
Fixture mode for deterministic tests; Playwright mode for real Chromium checks
Safety default
Private/local network targets blocked unless explicitly enabled for trusted local development
What This Measures
These checks measure BrowserOps as an agent-facing browser QA component: can an agent ask for a bounded page observation or verification, get structured evidence back, and hand that evidence to a reviewer without rerunning the browser.
What This Does Not Claim
This is not a claim that BrowserOps is a generic hosted browser farm, stealth scraper, CAPTCHA solver, or universal browser benchmark winner. The current claim is narrower: reliable browser evidence and UI verification for governed agent workflows.
Current Launch Gate
BrowserOps is ready to be evaluated as part of the Developer Bundle when the hosted route, production environment variables, Stripe catalog, and customer-facing scenario examples are wired into the launch app.
Sources and Scope
- BrowserOps deterministic suite: npm test covers observation, verification, contracts, MCP stdio, CLI flows, receipt validation, private-target refusal, and doctor diagnostics.
- Real browser smoke: optional Playwright observation and verification checks pass when BrowserOps is pointed at a working Chromium binary with BROWSEROPS_CHROMIUM_EXECUTABLE_PATH.
- Safety scope: BrowserOps blocks local/private network browsing by default and does not include credential theft, CAPTCHA bypass, stealth scraping, or unrestricted autonomous browsing.
BrowserOps research should stay reliability-scoped until there is a pinned public browser-workflow benchmark with comparable third-party baselines. For launch, this page documents the evidence path a buyer can inspect and reproduce.
Product Link
Use BrowserOps
The product page connects these reliability notes to the hosted MCP plan, Developer Bundle inclusion, and checkout path.