cdp-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| take_screenshotB | Capture a pixel-perfect fullpage or element-level screenshot of a webpage using real Chrome/Edge CDP without Puppeteer/Playwright dependencies. |
| interact_and_screenshotB | Simulate user input (with React/Vue controlled component penetration) and/or button click on a webpage, then capture a screenshot. |
| inspect_elementB | Inspect a DOM element on a webpage via CDP: check if it exists, its computed physical bounding box (x, y, width, height), and whether it is currently visible. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 3 tools
take_screenshot and interact_and_screenshot overlap since the latter is essentially a superset that also produces a screenshot, so an agent may be unsure whether to use the simpler tool or always reach for the interactive one. inspect_element is clearly distinct, and the descriptions do help clarify the intent boundary.
All names are snake_case and start with an action verb (take_, interact_, inspect_), which is mostly consistent. interact_and_screenshot deviates slightly by embedding a conjunction and a compound verb phrase rather than a clean verb_noun form.
Three tightly scoped tools is on the lean side but each earns its place for a focused CDP screenshot/inspection server. It stays well under the point where a small surface would feel arbitrary or padding-driven.
The surface covers capture, interaction, and element inspection, but notable gaps remain: no explicit navigation/URL loading tool, no wait-for-selector or network-idle readiness control, and no way to retrieve console/network diagnostics or script evaluation. Agents will hit dead ends for common flows like waiting for dynamic content before capturing.