browspark (uncaughterrs/browspark) is an MCP server listed on the M8ven Trust Index. It scores 52 out of 100, grade D. It declares 43 tools. No publisher has claimed this listing.
BrowserMCP and Chrome extension that let AI agents browse, click, and debug real websites in your own browser. Console, network, debugger, profiling, and Lighthouse, all from one tool call.
Caution. Specific findings reduced this grade. They are listed on the page. Grades reflect the full trust pyramid: code, verification depth, and reputation. New projects cap at C until adoption is earned.
How we verified
Verified is a snapshot. Live keeps it current, and builds your track record.
⚡ Connect GitHub → continuous verification on every pushwhy connect →Who stands behind it
uncaughterrs
Source: github_repo_search
Claim it to get a verified publisher badge, a free copy of our full audit findings, and direct contact for any high-priority issues we find. Or connect your repo for our deepest verification, Live Monitored: read-only, revoke anytime. What we access →
Install from
The grade above is for the source repository. Registries can serve a different version, so we mark the ones we were not able to read.
These names and descriptions are the publisher's own, read from the source code. We print them as written. Our assessment is the findings above, not this list.
browser_statusConnection status for both modes, how to get the extension connected, shared tabs across all windows, and open dialogs. Select a listed tabId to target a tab. Call this first if anything fails.
browser_sessionDeveloper-mode browsers: a separate Chrome the companion launches, without the user's logins. Only launch one when the user explicitly asks for it, or when a tool reported that an operation needs developer mode (heap snapshots, Lighthouse, raw CDP). For everyday work use the user's shared tabs. Each…
browser_tabsList, open, close, or activate tabs across all windows. Select a listed tabId to target a shared tab. Without tabId, tools use the agent's own usable tab or the only usable tab; multiple shared tabs require an explicit choice. Open a new tab only when the task needs a separate page. Never launch a s…
browser_navigateNavigate a tab: goto a URL, reload, back, or forward. Waits for the load event.
browser_snapshotAccessible snapshot of the page as an indented tree. Interactive elements get refs like [ref=e12] for click/fill/select/read/wait and the devtools_elements tool. Refs stay valid until the DOM changes; re-snapshot after navigation. diff:true returns only lines that changed since the previous snapshot…
browser_readRead page content: visible text, markdown (main content converted with headings, lists, links, code, tables), links, tables (as row arrays), or raw HTML. Optionally scoped to a ref.
browser_fetchFetch a URL and return its main content as markdown (or text/html) without leaving a tab behind: opens a background agent tab, waits for load, reads, closes. Use for reading docs and articles.
browser_policyRestrict where a tab (or every tab the agent opens from now on) may navigate and load from: allow lists domains that are permitted (everything else is blocked), block lists domains that are denied. Subdomains match. Non-http schemes always pass. Blocked navigations fail with ERR_BLOCKED_BY_CLIENT.
browser_downloadDownloads started in the browser: list them, or wait for the newest matching download to complete, including one already completed, and get its file path when available. Developer mode saves into the context's download directory; extension mode reports downloads observed on shared tabs.
browser_pdfPrint the page to a PDF file (like Print → Save as PDF). Returns the file path.
browser_webmcpWebMCP: pages can expose typed tools to agents via navigator.modelContext. list reports what the page exposes (or that it exposes nothing); invoke calls one by name.
browser_extractStructured extraction: for every element matching `items`, read named fields as text or an attribute. Returns an array of objects.
browser_clickClick an element by ref with real mouse events. Also hover (hover:true) and drag (dragTo).
browser_fillReplace the contents of a text input, textarea, or contenteditable. Use browser_select for <select>, browser_click for checkboxes, browser_upload for files.
browser_selectChoose option(s) in a <select> by value or visible label.
browser_uploadSet files on an <input type=file> by ref. Paths must be readable by the browser machine.
browser_keybrowser_scrollScroll the page, or the nearest scrollable ancestor of a ref.
browser_screenshotScreenshot the viewport, the full page, or one element.
browser_waitWait until text appears/disappears, the URL contains a string, or a ref reaches a state. Polls up to timeoutMs (default 15000).
browser_dialogAccept or dismiss an open JavaScript alert/confirm/prompt/beforeunload dialog. browser_status shows open dialogs.
browser_batchRun several tool calls in order in one round trip. Stops at the first error. Each step is {tool, args}.
devtools_storageApplication storage. area: cookies (list/set/delete), local or session storage (list/set/remove/clear), indexeddb (databases/stores/data/put/clearStore/deleteDatabase), cache (list/entries/delete), usage (quota and per-type usage), clear (explicitly named storage types for one origin). Origin defaul…
devtools_workersService workers and PWA. list registrations, versions, and worker targets; update, unregister, skipWaiting, start, stop a worker; bypass routes requests around service workers; forceUpdate reloads the worker on every navigation; manifest reads the web app manifest and installability errors; evaluate…
devtools_consoleSearch collected console output: logs, warnings, errors, uncaught exceptions, and browser log entries with stack traces. Preserved across reloads while the session runs. Actions: search (default), get (one message with full stack), inspect (an object logged by id/objectId), wait (block until a match…
devtools_evaluateEvaluate JavaScript in the page (or a specific frame/context, or the paused call frame) and return the value or the exception. Objects come back as previews with an objectId for devtools_console inspect. awaitPromise defaults to true.
devtools_debuggerJavaScript debugger. Breakpoints: set (by url/scriptId or original source via source map; line 1-based; optional condition or logMessage), remove, enable, disable, list; exceptions (none|uncaught|all); dom (subtree-modified|attribute-modified|node-removed on a ref/selector); event (listener breakpoi…
devtools_elementsElements panel. Target an element by ref (from browser_snapshot), CSS selector, or nodeId. Actions: search (CSS selector, XPath, or text; returns refs), describe, html (get/set outerHTML), attributes (get/set/remove), classes (add/remove/toggle), styles (matched rules + inline; setStyle edits inline…
devtools_emulationDevice and rendering emulation. device (preset or custom viewport, scale, touch, user agent), viewport, cpu (throttle rate), network (preset), geolocation, media (color scheme, reduced motion, forced colors, print), locale/timezone, vision (deficiency), animations (playback rate, list observed), sta…
devtools_accessibilityAccessibility tree and browser-reported issues. tree returns roles, names, and states for the page or a subtree (by ref/selector); node returns full properties for one element; check runs a quick audit (missing alt/labels/names, duplicate ids); issues lists problems Chrome reported (Audits domain: c…
devtools_securitySecurity panel. state: connection security, certificate (subject, issuer, validity, protocol, cipher), and Chrome security issue ids. certificate: details for the main document or a URL. mixedContent: insecure subresources on a secure page. issues: security-related browser-reported issues (mixed con…
devtools_lighthouseRun Lighthouse (official CLI) against the developer-mode browser and return category scores, key metrics, failing audits, plus HTML and JSON report artifacts. Developer mode only; Lighthouse opens its own tab in that browser.
devtools_networkInspect network activity collected in the session (preserved across reloads). Actions: search (filters below; searchIn picks url/headers/body/postData), get (one request: headers, timing, initiator stack, redirects, security), body (response body, JSON pretty-printed; fetched on demand if not captur…
devtools_performancePerformance panel. start/stop a trace recording (Chrome trace format, opens in DevTools Performance and Perfetto). stop returns a summary: time by category, long tasks with their heaviest children, script time by URL, observed LCP/FCP/CLS, user timings. search finds trace events by name/args. compar…
devtools_profileCPU profiler. start/stop a sampling profile (.cpuprofile artifact, opens in DevTools Performance/JavaScript Profiler). stop and analyze return a bottom-up list of expensive functions (self time) and a call tree of hot paths.
devtools_memoryMemory panel. snapshot captures a .heapsnapshot artifact (opens in DevTools Memory) and summarizes objects by class with sizes and detached DOM node count. compare diffs two snapshots by class (growth). retainers lists what holds instances of a class. sampling start/stop records an allocation sampli…
devtools_coverageCoverage panel. start records JavaScript and CSS usage; stop reports per-file used/unused bytes; detail lists the unused ranges of one file with line numbers and snippets.
devtools_recorderRecord and replay browser actions. start records every browser_navigate/click/fill/select/key/wait on the tab (with a stable selector per element); stop saves the flow as a JSON artifact. add_assertion appends a browser_wait check; add_checkpoint marks a place replay can stop at; parameterize turns …
devtools_sessionStart or stop an inspection session on a tab. While active, console, network, exceptions, issues, scripts, and debugger events are collected across reloads. Stop removes every breakpoint, override, and emulation the session set. clear drops collected data.
devtools_eventsRead or wait for events collected in the inspection session: console messages, exceptions, request start/finish/fail, breakpoint pauses, navigations, issues, service worker and animation events, recording completion. Use afterId to page forward; wait:true blocks until a matching event arrives.
devtools_capabilitiesReport the browser version, connection mode, and which CDP domains work for a tab (probed live, cached per browser). Extension mode blocks some domains that direct CDP allows.
devtools_cdpDeveloper mode only: run any Chrome DevTools Protocol command on a tab (or on the browser target with target:"browser"). Params must be a JSON object valid for the method.
devtools_sourcesSources panel: list loaded scripts, stylesheets, documents, frames and execution contexts; get file contents (optionally the original source via source map); search across sources by filename, text, or regex; resolve source maps; apply, list, and revert temporary resource overrides (served in place …
Disclosed vulnerabilities in this server's declared npm dependencies (via OSV). Whether each is reachable depends on the installed versions.
@modelcontextprotocol/sdk has cross-client data leak via shared server/transport instance reuse
Anthropic's MCP TypeScript SDK has a ReDoS vulnerability
Model Context Protocol (MCP) TypeScript SDK does not enable DNS rebinding protection by default
ws: Memory exhaustion DoS from tiny fragments and data chunks
ws: Uninitialized memory disclosure
BROWSPARK_ARTIFACTSBROWSPARK_CHROMEBROWSPARK_PORTBROWSPARK_PROFILEBROWSPARK_PROFILESTool annotations
No tools have read-only/destructive annotations
Add readOnlyHint or destructiveHint annotations to every tool so hosts can warn users before invoking.
All four hints declared on every tool
43/43 tools missing one or more hints — browser_status (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); browser_session (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); browser_tabs (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint), +40 more. OpenAI's directory rejects tools where any of the four hints are missing or non-boolean.
For every tool, set all four hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) to explicit true/false values that match the handler’s actual behaviour.
Production dependencies are patched
0 critical, 4 high severity in production deps — @modelcontextprotocol/sdk@1.20.0 (high), @modelcontextprotocol/sdk@1.20.0 (high)
Run npm audit fix, or upgrade the affected packages to a non-vulnerable version.
Tool description accuracy
devtools_recorder: description implies read-only but handler writes/deletes/executes
Update tool descriptions to accurately reflect all capabilities — especially write, delete, or execute operations.
Claim the listing to review these findings one by one and send us a correction where you disagree, straight to the team. Claiming also means we tell you when the grade moves, and reach you first if we find anything urgent.
[](https://m8ven.ai/mcp/uncaughterrs/browspark)?variant=verified from the URL.Vetting this one by hand? Tool Check is an MCP that scores other MCPs. Add it once and ask Claude, ChatGPT, or any MCP client to grade a server, surface CVEs, check the publisher, and suggest safer alternatives — before you install.
https://m8ven.ai/api/mcp/tool-check