Skip to main content
Glama

caido-screenshot-mcp

An MCP server that captures formatted screenshots of Caido Replay request/response pairs, by driving the Screenshot Mode plugin.

Caido has no screenshot API — Screenshot Mode is a frontend-only plugin that renders the DOM to a PNG in the browser. This server reaches it over the Chrome DevTools Protocol, clicks the plugin's own export, and intercepts the resulting image before Electron can open a save dialog. The output is the plugin's real output, written straight to disk.

Requirements

  • Node >= 22 (uses the built-in WebSocket; there are no npm dependencies)

  • Caido desktop

  • The Screenshot Mode plugin installed in Caido

  • Caido launched with --remote-debugging-port (see below)

Related MCP server: MCP Browser Screenshot Server

Install

git clone <this-repo> caido-screenshot-mcp
cd caido-screenshot-mcp
npm link            # optional: puts `caido-screenshot` on your PATH

No npm install step — there are no dependencies.

Start Caido with the debug port

The plugin renders inside Caido's window, so the server needs to talk to that window. Launching Caido normally leaves the port closed and every call will fail.

caido-screenshot launch          # quits nothing; starts Caido with the port open

Or manually:

# macOS
open -a Caido --args --remote-debugging-port=9222
# Linux
caido --remote-debugging-port=9222
# Windows
"%LOCALAPPDATA%\Programs\Caido\Caido.exe" --remote-debugging-port=9222

Check everything at once:

caido-screenshot doctor

Use with Claude Code

claude mcp add caido-screenshot --scope user -- node /absolute/path/to/caido-screenshot-mcp/src/server.mjs

Restart Claude Code. Then just ask: "screenshot the current Caido replay session".

Use with any MCP client

{
  "mcpServers": {
    "caido-screenshot": {
      "command": "node",
      "args": ["/absolute/path/to/caido-screenshot-mcp/src/server.mjs"]
    }
  }
}

Tool: caido_screenshot

Parameter

Type

Description

out_path

string

Where to save the PNG. Default ~/Downloads/caido-screenshot-<timestamp>.png

collection

string

Replay collection containing the session. It is expanded automatically.

session

string

Replay session to select first, by sidebar name (e.g. GET /login)

session_id

string

Exact Replay session ID. Preferred because it is unambiguous.

require_response

boolean

Reject request-only sessions and stale previews. Default true.

inline

boolean

Also return the image in the result (large). Default false

Examples

Once the server is registered you never name the tool. Describe what you want in plain English and the agent picks it up.

Just capture what's on screen

screenshot the current Caido replay request

The agent calls caido_screenshot and replies with the saved path:

Saved /Users/you/Downloads/caido-screenshot-2026-08-22T13-40-11-002Z.png (438 KB)

Send a request, then capture it

This is the common one. Pair it with the official Caido MCP server, which can send requests — this server captures the result.

send a GET to https://httpbin.org/get in Caido replay and screenshot it

The agent sends the request, maps its returned Replay entry to the populated session, and captures that exact session ID:

Step

Tool

Server

1

caido_send_request

caido

2

caido_list_replay_sessions

caido

3

caido_screenshot(session_id: ...)

caido-screenshot

Match caido_send_request.entryId to caido_list_replay_sessions[].activeEntryId. Do not create a new session from requestSourceId for response evidence: Caido seeds the request template but does not copy the completed response.

Send a raw request you wrote

Paste the raw HTTP and say what you want:

send this in Caido replay and screenshot it:

POST /post HTTP/1.1
Host: httpbin.org
Content-Type: application/json

{"user":"demo","token":"abc123"}

Pick a specific session

Sessions are named after the request line in Caido's sidebar, so GET /, POST /login and so on:

screenshot the "POST /login" replay session

That becomes caido_screenshot(session: "POST /login"). Every capture uses Side by Side.

For reliable automation, use the session ID:

caido_screenshot(session_id: "67", require_response: true)

If the session is inside a different or collapsed collection, name both:

screenshot "01 admin create private conversation" from the "OBS AI Assistant IDOR - Fresh Evidence" collection

That becomes caido_screenshot(collection: "OBS AI Assistant IDOR - Fresh Evidence", session: "01 admin create private conversation"). The tool expands the collection before selecting the session and captures it Side by Side.

Heads up: caido_send_request auto-names sessions from the request line, so several targets can all end up called GET /. Rename a session in Caido before capturing if you need to refer to it unambiguously.

Save somewhere specific

screenshot it and save to ~/reports/login-flow.png

Using it from Codex

Start an interactive session and ask the same way:

codex
> screenshot the current Caido replay session

Codex asks for approval the first time. To skip the prompt for this tool, add to ~/.codex/config.toml:

[mcp_servers.caido-screenshot.tools.caido_screenshot]
approval_mode = "auto"   # auto | prompt | writes | approve

codex exec (non-interactive) will not work — it forces approval: never and refuses all MCP tool calls, regardless of per-tool settings. Use an interactive session.

CLI

caido-screenshot                                   # current session -> ~/Downloads
caido-screenshot shot.png                          # explicit path
caido-screenshot shot.png --session "GET /login"
caido-screenshot shot.png --session-id 67
caido-screenshot shot.png --collection "Fresh Evidence" --session "01 admin request"
caido-screenshot doctor
caido-screenshot launch

Environment

Variable

Default

Description

CAIDO_CDP_PORT

9222

Caido's remote debugging port

CAIDO_CDP_TIMEOUT_MS

45000

Per-call CDP timeout

Known limitations

These are properties of the plugin, not of this server:

  • Long content is cut off. Caido's editors are CodeMirror, which only renders the visible lines. Anything taller than the Caido window is not in the DOM and cannot be captured. Very long responses can come out blank; the server errors instead of writing a near-empty file.

  • Headers can be dropped. The DOM-to-image render sometimes omits lines, with visible gaps in the line numbers. Treat these images as illustrative, not as byte-exact evidence — use Caido's own request export when precision matters.

  • The layout is fixed to Side by Side. The server deliberately exposes no layout or arrangement option so every evidence screenshot has the same report-ready format.

  • Screenshots are not redacted. Whatever is on screen — Authorization, Cookie, Set-Cookie — is captured in full. Review before sharing.

  • Request-only Replay sessions are rejected by default. Screenshot Mode can retain its previous preview when the selected session has no response. Pass --allow-no-response only when a request-only screenshot is intentional.

Troubleshooting

Message

Cause

Cannot reach Caido's debug port

Caido not launched with --remote-debugging-port. Quit it and run caido-screenshot launch.

Caido is showing the instance picker

Click Connect on your local instance in the Caido window.

No Screenshot button in the Replay toolbar

No Replay session open, or the Screenshot Mode plugin isn't installed/enabled.

No Replay collection matching "..."

Use the exact collection name shown in the Replay sidebar.

No Replay session matching "..."

The tool expands collection when supplied, then lists the visible session names if no exact match appears.

has no rendered response

The selected session was seeded from a request or has not completed. Capture the session whose activeEntryId equals the send result's entryId.

preview is stale

Screenshot Mode showed a different host/path (or a conflicting visible query). Nothing was saved; retry using the populated session_id. Scheme, encoding, query order, omitted query labels, and Screenshot Mode 0.8.1's missing ? delimiter are normalized automatically.

returned an empty image

Content too long to render. Capture a shorter request/response.

License

MIT

Available Tools

1 tool
caido_screenshotA

Capture a formatted screenshot of a Caido Replay request/response using the Screenshot Mode plugin. Requires Caido running with --remote-debugging-port (default 9222) and a Replay session open. Uses one consistent Side by Side layout, verifies the selected session, requires a rendered response by default, rejects stale previews, then saves a PNG and returns its path plus the verified request metadata.

ParametersJSON Schema
NameRequiredDescriptionDefault
inlineNoAlso return the image in the tool result (large). Default false.
sessionNoReplay session to select first, by the name shown in the sidebar (e.g. 'GET /login')
out_pathNoWhere to save the PNG. Defaults to ~/Downloads/caido-screenshot-<timestamp>.png
collectionNoReplay collection containing the session. The collection is expanded automatically.
session_idNoExact Replay session ID. Preferred over session name because it is unambiguous.
require_responseNoReject sessions without a rendered response to prevent stale evidence. Default true.

TDQS

A4.5/5.0
Behavior5/5

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

With no annotations, the description carries the full burden of transparency. It discloses multiple behavioral traits: uses a specific layout, verifies the selected session, requires a rendered response by default, rejects stale previews, saves a PNG, and returns path plus metadata. This gives the agent a strong understanding of what happens during execution.

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

Conciseness5/5

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

The description is concise and well-structured, opening with the core action and then efficiently enumerating prerequisites and key behaviors. It avoids unnecessary verbosity while covering all essential aspects, making it easy for an agent to parse quickly.

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 complexity of the tool (multiple preconditions, specific layout, output behavior), the description provides sufficient context for an agent to decide when to use it. It clearly states the environment requirements, what is verified, and what is returned, without needing an output schema. The description is complete enough for operational use.

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

Parameters3/5

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

The schema provides 100% coverage with detailed descriptions for each parameter, so the baseline is 3. The tool description does not add additional insight beyond the schema's parameter explanations; it merely references the default behavior of require_response without expanding on relationships or interdependencies.

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 captures a formatted screenshot of a Caido Replay request/response using the Screenshot Mode plugin, specifying the exact resource and action. It also distinguishes itself by mentioning the Side by Side layout and the plugin, which sets it apart from generic screenshot tools.

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

Usage Guidelines4/5

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

The description provides explicit preconditions for use (Caido running with --remote-debugging-port, Replay session open) and mentions key behaviors (requires rendered response, rejects stale previews). While it doesn't explicitly contrast with alternatives, the specific prerequisites and layout make the intended use case clear.

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

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 1 tool updatev1.2.1
    • First observedcaido_screenshot

TDQS

A4.4/5.0

Scored across 1 tool

Disambiguation5/5

With only one tool, there is no possibility of confusion between tools. The tool has a single, clearly defined purpose (capturing a screenshot of a Caido Replay session), so an agent cannot misselect it.

Naming Consistency5/5

The single tool follows a consistent snake_case naming convention (caido_screenshot), which is clear and predictable. There is no mix of styles or verbs to create inconsistency.

Tool Count3/5

The server provides only one tool, which feels thin for a typical MCP server. However, the tool targets a very specific, narrow use case (Caido Replay screenshot capture), so the count may be appropriate for that focused purpose, but it is borderline.

Completeness4/5

The tool covers the core workflow of capturing a formatted screenshot from a Replay session, including verification and metadata return. There are minor gaps (e.g., no options for custom layout or output format), but for its stated purpose, it is essentially complete.

Maintenance

ActivityMaintained
ResponsivenessNo issues

Related MCP Connectors

Related MCP Servers

  • A
    license
    A
    quality
    A
    maintenance
    Enables capturing screenshots of web pages and local HTML files through a simple MCP tool interface using Puppeteer with configurable options for dimensions and output paths.
    2
    464 npm
    30
    Apache 2.0
  • A
    license
    C
    quality
    D
    maintenance
    Enables taking screenshots of web pages with support for multiple devices (desktop, mobile, tablet), custom dimensions, full-page capture, and various image formats. Built with Playwright for reliable web page rendering and screenshot generation.
    1
    MIT