Skip to main content
Glama

browser_eval

Inspect or manipulate a browser tab by evaluating JavaScript, retrieving HTML, or extracting server-rendered SPA state from Next, Nuxt, Remix, Apollo, GitHub, or Redux frameworks.

Instructions

Purpose: Inspect or operate on a browser tab via 3 actions: 'js' (evaluate JS), 'dom' (get HTML), 'appState' (extract SSR-injected SPA state). Details: action='js' — Run a JS expression. withPerception:true wraps in {ok, result, post}. action='dom' — Return outerHTML of selector (or document.body), truncated to maxLength. action='appState' — Scan Next/Nuxt/Remix/Apollo/GitHub/Redux SSR injected JSON; pass selectors to override defaults. Prefer: Use action='appState' BEFORE 'dom' or 'js' on SPAs where rendered HTML is sparse — single CDP call. Use 'dom' when 'appState' is empty and you need page structure. Use 'js' as the escape hatch for arbitrary scripting. Caveats: DOM nodes cannot be returned from action='js' directly (circular refs are serialized safely). React/Vue/Svelte controlled inputs cannot be set via element.value — use keyboard(action='type') / browser_fill instead. readyState is strictly checked; guard blocks if page is still loading. Typed errors: code:'BrowserNotConnected' on CDP disconnect (re-attach via browser_open); code:'AutoGuardBlocked' when the auto-guard refuses (e.g. page still loading) — the error message preserves the guard's 1-sentence recommended next step (most often wait_until({condition:'ready_state'}) or browser_eval readyState polling, then retry). Examples: browser_eval({action:'js', expression:'document.title'}) → page title browser_eval({action:'dom', selector:'#main', maxLength:5000}) → outerHTML browser_eval({action:'appState'}) → default SPA state probes

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
portNoChrome/Edge CDP remote debugging port.
tabIdNoTab ID from browser_open. Omit to use the first page tab.
actionYesAction selector — one of: js, dom, appState. Per-action required fields are enforced at call time (see the tool description); this flat schema lists every action's fields as optional.
lensIdNoOptional perception lens ID. Guards (target.identityStable) are evaluated before eval.
includeNoOptional response-shape opt-in. `['envelope']` returns the self-documenting envelope (`_version` / `data` / `as_of` / `confidence`). `['raw']` forces raw shape (overrides DESKTOP_TOUCH_ENVELOPE=1 server default). Default behaviour is raw shape (compat with existing clients).
maxBytesNoMax bytes per individual payload (default 4000). Larger payloads are truncated.
selectorNoCSS selector for root element. Omit for document.body.
maxLengthNoMax characters of HTML to return (default 10000).
selectorsNoCustom probe selectors. Omit to use the default SPA framework set (__NEXT_DATA__ / __NUXT_DATA__ / __REMIX_CONTEXT__ / __APOLLO_STATE__ / window:__INITIAL_STATE__ etc.). Window globals must be prefixed with 'window:'.
expressionNoJavaScript expression to evaluate. The server automatically wraps snippets in an async IIFE to avoid repeated const/let collisions. For multi-statement snippets, use an explicit final return value. Declarations (const/let/var) are scoped per snippet — use window.* / globalThis.* for persistence. A single eval is bounded by the CDP per-command timeout (~15s): do NOT write in-page polling loops here — use wait_until (element_matches / url_matches / ready_state) to wait for conditions instead.
includeContextNoWhen true, append activeTab and readyState context to the response.
withPerceptionNoWhen true, return structured JSON {ok, result, post} with post.perception attached. Default false preserves raw-text return.
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description bears full responsibility for disclosing behavior. It details error types (e.g., 'BrowserNotConnected', 'AutoGuardBlocked'), truncation behavior, readyState checks, and limitations like circular reference serialization. This fully informs the agent of the tool's behavior.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is lengthy but well-structured with clear sections (Purpose, Details, Prefer, Caveats, Examples). It is front-loaded with the essential purpose and action list. While some redundancy could be trimmed, the structure and content are appropriate for the tool's complexity.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (12 parameters, no output schema), the description is thorough. It covers all actions, usage preferences, caveats, error handling, and provides examples. It addresses potential edge cases and integrates with sibling tools, ensuring completeness.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so baseline is 3. The description adds significant value beyond the schema by explaining parameter interactions (e.g., 'withPerception' returns structured JSON, 'includeContext' appends context) and providing examples. It enhances understanding of parameter semantics.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose: inspect or operate on a browser tab via three distinct actions ('js', 'dom', 'appState'). It differentiates from sibling tools by specifying when to use each action and provides examples, making the purpose very clear.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides explicit usage guidelines, including a 'Prefer' section that advises when to use each action (e.g., use 'appState' before 'dom' or 'js' on SPAs). It also includes caveats for when not to use certain actions, such as avoiding direct DOM node return from 'js'. This thoroughly guides the agent.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/wangkang1133/desktop-touch-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server