Skip to main content
Glama

Robot Actions — Remote Device Control

android_devtools_capture_network

Capture network requests made by a page running on the device over a time window via CDP. Enables the Network domain and collects request/response/loadingFinished/loadingFailed events. LIVE-WINDOW collector: only captures requests made AFTER it attaches — to capture a navigation/reload/fetch, pass triggerJs (fired inside the window) rather than triggering it from a separate tool call (which races the attach and returns 0 records). Returns { records, bodiesOmitted? } in summary format, or a HAR 1.2 document when format="har". Each record: { requestId, method, url, requestHeaders?, status?, statusText?, mimeType?, resourceType?, responseHeaders?, encodedDataLength?, state, errorText?, startTimestamp?, endTimestamp?, body?, bodyTruncated?, bodyError? }. Set includeBodies=true to fetch response bodies for completed text-like responses (json|text|xml|javascript|html|css|svg|x-www-form-urlencoded); per-body cap: 10 000 chars (bodyTruncated=true when hit); total cap: 200 000 chars (excess records counted in bodiesOmitted). Body fetch failures set bodyError on that record. Set throttle to emulate network conditions for the capture window only (cleared in a finally block): slow-3g (latency 2000ms, 400kbps down/up), fast-3g (latency 563ms, 1.6Mbps down / 750kbps up), offline. When throttle was active, a top-level throttle field appears in the output. Set blockUrls (URL patterns) to block matching requests, or extraHeaders to inject request headers (e.g. Authorization), for the capture window only — both cleared afterward. Returns at most limit records (default 100, most-recent first) so heavy pages stay within the token budget — filter with urlSubstring / onlyErrors; total/returned appear when records were dropped. Default window: 5 000 ms. Maximum: 30 000 ms. Omit pageId to auto-select the visible/active page.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
udidYesDevice serial number (UDID)
limitNoMax records to return, most-recent first (default 100). Heavy pages make hundreds of requests; the default keeps the response within the MCP token budget. total/returned are included when records were dropped.
formatNoOutput format: "summary" (default) returns { records, bodiesOmitted? }; "har" returns a HAR 1.2 document
pageIdNoTarget page id (auto-picked when omitted)
socketNoAbstract unix socket name (default: chrome_devtools_remote)
throttleNoEmulate network conditions for the capture window only (cleared in a finally block). slow-3g: latency 2000ms, 400kbps down/up. fast-3g: latency 563ms, 1.6Mbps down / 750kbps up. offline: no connectivity.
blockUrlsNoBlock requests whose URL matches any of these patterns (wildcards allowed, e.g. "*.doubleclick.net", "*/analytics*") for the capture window ONLY (cleared afterward). Blocked requests surface as failed. Use to capture without third-party/analytics noise.
triggerJsNoJavaScript run INSIDE the capture window (right after the collector attaches) so the requests it triggers are captured race-free — e.g. "location.reload()", "location.href='https://example.com'", or "fetch('/api/x')". Prefer this over navigating/reloading from a separate tool call, which races the collector attach and yields 0 records.
durationMsNoCollection window in milliseconds (default: 5000, max: 30000)
onlyErrorsNoOnly failed / 5xx records
extraHeadersNoExtra HTTP headers added to every request for the capture window ONLY (cleared afterward), e.g. { "Authorization": "Bearer …" } to capture authenticated requests. These are session-scoped, which is why they apply only during this capture.
urlSubstringNoOnly records whose URL contains this substring — target the requests you care about
includeBodiesNoFetch response body for completed text-like responses (default: true). Request bodies (postData) are always captured. Set false to skip response bodies for a lighter payload. Per-body cap: 10 000 chars. Total cap: 200 000 chars.

TDQS

A4.8/5.0
Behavior5/5

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

No annotations are provided, so the description carries the full behavioral burden and it delivers richly. It discloses the live-window race condition, the exactly-once cleanup semantics (throttle/blockUrls/extraHeaders cleared in a finally block), return format details (records vs HAR, per-body 10000-char cap, total 200000-char cap with bodiesOmitted), the limit-based record capping behavior (total/returned when dropped), and the throttle semantics including exact latency/bandwidth figures. This is far more transparent than the typical tool.

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 front-loaded with the core purpose and key caveat (live-window), then flows logically through return format, body handling, throttle, blocking, and limits. Every sentence adds operational value. It's a long single paragraph, so it loses a point for not breaking into scannable sections, but given the tool's complexity (13 params, multiple modes, many caps) the length is justified and there's no waste.

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?

For a 13-parameter tool with no output schema and no annotations, this description is remarkably complete. It documents return record shape, all caps (per-body, total, limit), the throttle values with exact figures, blockUrls wildcard semantics, extraHeaders session-scoping rationale, and the format="har" alternative. There is no output schema so the description must convey the return shape, and it does so thoroughly. This is near the ceiling for what a description can achieve.

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 description coverage is 100%, so the baseline is 3. The description adds genuine value beyond the schema: it explains the interplay between triggerJs and the live window (race avoidance), clarifes that includeBodies has a default of true while the schema doesn't state a default, and explains the bodiesOmitted mechanism tied to total caps. The main shortcoming is that with 13 params the description cannot deep-dive each, but the schema itself is already rich so this is acceptable.

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 opens with a specific verb+resource+scope: 'Capture network requests made by a page running on the device over a time window via CDP.' It clearly states it enables the Network domain and collects specific event types (request/response/loadingFinished/loadingFailed). This is immediately differentiated from siblings like device_network_info (device-level info) and ios_safari_capture_network (iOS targeting) via the 'device' and page-based (CDP) framing.

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?

Exceptionally explicit on when/how to use: it distinguishes the LIVE-WINDOW collector behavior and instructs to pass triggerJs inside the window rather than triggering from a separate tool call to avoid racing the attach and getting 0 records. It also names the exact default window (5000ms), maximum (30000ms), and the pageId auto-selection note. The triggerJs description even gives concrete examples like "location.reload()" and "fetch('/api/x')", which is strong operational guidance.

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

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

B3.1/5.0
Disambiguation2/5

The set contains near-identical duplicate families: web_* and playwright_* expose ~15 pairs of the same desktop-grid-browser operations (web_get_text/playwright_get_text, web_reload/playwright_reload), and screenshot/log/network/mock capabilities each have 5-8 entry points (device_screenshot vs android_mjpeg_screenshot vs ios_screenshot vs ios_fast_screenshot vs web_screenshot vs webpage_screenshot vs session_screenshot). Many individual descriptions carefully draw boundaries (devtools vs traffic, HID vs session), but an agent cannot reliably distinguish web_* from playwright_*, and ios_screenshot/ios_fast_screenshot/ios_mjpeg_screenshot blur together.

Naming Consistency2/5

The prefix scheme is broken: Android functionality is split arbitrarily between android_* and device_* (device_screenshot vs android_mjpeg_screenshot), the desktop browser gets two parallel prefixes (web_* and playwright_*), and verbs vary across equivalents (device_navigate_url vs web_navigate vs ios_safari_navigate). session_* uses bare verbs (session_url, session_back), and the same concept gets different names (ios_clipboard_get_hid vs ios_get_pasteboard; device_screen vs ios_orientation).

Tool Count1/5

333 tools is an extreme count by any measure — far beyond the 50+ threshold — and much of the bulk is duplicative (the web_*/playwright_* pairs alone double ~15 slots) or out-of-scope for a device-control server (TestRail, Jira, AzDO, agent memory, secret variables, feedback). Even granting that remote device control + test automation is a broad domain, this surface will devastate agent context budgets and is impossible to navigate coherently.

Completeness4/5

The core device-control and test-automation domain is remarkably thorough: Android and iOS each have full interaction, app-lifecycle, file, network/proxy, performance, crash, accessibility, recording, and replay coverage, with CRUD lifecycles for flows, suites, app uploads, TestRail cases, and visual-review baselines. Minor gaps exist at the margins — Jira/AzDO lack update/transition/comment operations, and iOS cannot open/close tabs — but the central workflows have no dead ends.

Resources