M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
A

Anything Analyzer MCP Server

Community
Cross-scenario traffic capture and protocol analysis for AI agents.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #422 of 438
Stars
★ 3.8k Very popular
Runtime
Prebuilt binary
Credentials
No credential needed
License
Apache-2.0
Last commit
27FMRS · D

The README explicitly describes a built-in MCP Server, an MCP Client, browser capture, MITM proxying, and AI analysis, and names Claude Desktop and Cursor. Because no server manifest, tool names, transport details, or client configuration example is supplied, the MCP interface cannot be independently verified here. Repository metadata marks the license as unknown, while the README separately displays an MIT license badge and states MIT in its license section.

Strongest · Documentation 9/20 Weakest · Reliability 3/20

Reliability
3/20
Security and permissions
5/20
Maintenance
4/20
Documentation
9/20
Setup experience
6/20
Why each score
Reliability 3/20
No server manifest, source files, CI, or test evidence is provided, so MCP initialization, tool-list handshake, tool behavior, and error handling cannot be verified. The README only claims that an embedded MCP Server exists, leaving key paths unreviewable; this warrants a very low score.
Security and permissions 5/20
The product handles HTTPS MITM, CA installation, system-proxy changes, cookie/storage snapshots, complete request and response capture, and transmission to a user-configured LLM. The README claims read-only capture, but there is no source evidence for privilege boundaries, redaction, exfiltration scope, or confirmation of sensitive operations. No explicit malicious behavior is evidenced, so the red-line rule is not triggered, but substantial points are deducted.
Maintenance 4/20
Only repository metadata and README material are available. There is no reviewable evidence of commit cadence, releases, dependency updates, issue response, security reporting, or maintainer governance. The README states MIT while the supplied repository metadata says the license is unknown, creating an evidence inconsistency and justifying a low score.
Documentation 9/20
The README includes multi-platform build instructions, proxy setup, certificate details, feature descriptions, and disclosed limits such as the body-size cap and lack of WebSocket decryption. However, it omits a reviewable MCP manifest, tool names and parameters, client configuration, authentication boundaries, troubleshooting, cost guidance, and detailed data-flow disclosure, so the score remains modest.
Setup experience 6/20
Application installation and source-build steps are fairly detailed, but no mainstream MCP-client configuration example is provided, and there is no evidence of an independent MCP Server launch command, transport configuration, or runtime requirements. CA installation, system-proxy setup, and LLM API-key configuration also add manual steps; static evidence is insufficient for a higher score.

Static review · not runListed 2026-08-14

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

  • Developers who need one workflow for browser, application, terminal, and mobile traffic
  • Researchers conducting authorized protocol reverse engineering, API debugging, or security audits
  • Users who want an AI agent to help interpret HTTP traffic

Not for

  • Unauthorized intrusion, data acquisition, or system control
  • Active proxy workflows that require modifying requests or responses
  • Use cases requiring decrypted WebSocket contents
  • Users seeking a built-in AI model without configuring an external model API

Required permissions

  • AI analysis requires an API key for OpenAI, Anthropic, or a compatible API
  • HTTPS MITM capture requires installing and managing a local CA certificate
  • Changing system proxy settings may require operating-system permissions
  • The proxy can access request and response headers and bodies, plus Cookie, localStorage, and sessionStorage snapshots

Risks and side effects

  • Captured traffic may contain tokens, cookies, credentials, or other sensitive data
  • Installing a local CA certificate and enabling a system proxy increases the visibility of local traffic
  • Sending captured data to an external LLM service can create privacy and compliance risks
  • Bodies are limited to 1 MB and binary content is skipped; WebSocket traffic is tunneled rather than decrypted

Setup

Before you start

Runtime：Prebuilt binary

Download the platform installer from GitHub Releases, or build from source. Before using AI analysis, configure an API key under Settings → LLM. Before capturing HTTPS traffic through the MITM proxy, install the CA certificate under Settings → MITM Proxy and enable the proxy; the default port is 8888. The specific installation configuration for the MCP Server is not documented in the supplied source.

Check that it works

Check that the built-in MCP Server from anything-analyzer appears in your MCP client's tool list (e.g. Claude Desktop or Cursor), then ask it to read the current session's captured requests; a returned request list confirms the connection.

Troubleshooting

  1. Confirm that a valid API key and model service are configured under Settings → LLM
  2. If HTTPS capture fails, verify that the CA certificate is installed and the client proxy points to 127.0.0.1:8888
  3. If a phone, desktop application, CLI, or script produces no traffic, check its proxy configuration or system proxy state
  4. For source builds, confirm Node.js >= 18, pnpm, and Visual Studio Build Tools on Windows where applicable
  5. If MCP does not connect, inspect the client configuration; the source does not specify the built-in server's transport or configuration format

Things to try

Once connected, you can ask your AI assistant things like:

  • Use the MCP capture tool to list all HTTP requests captured in the current session
  • Analyze the latest captured requests in this session and reverse-engineer the target site's API
  • Start a MITM capture via MCP and show the traffic from my curl requests on port 8888
  • Run a JS crypto reverse-engineering analysis on this session to identify encryption calls

Use cases

Reverse-engineering website APIs, authentication flows, and frontend encryption
Analyzing network protocols used by mobile apps, desktop applications, CLIs, and scripts
Performing security audits, performance analysis, and request-flow analysis
Connecting traffic capture and AI analysis to MCP clients

Supported clients

Claude DesktopPartial support
CursorPartial support

Listed from the project's documentation, not tested by this site.

Overview

Anything Analyzer is an Electron desktop tool that unifies traffic captured from its embedded browser and MITM proxy, then uses configurable OpenAI, Anthropic, or compatible LLM APIs for protocol reverse engineering, security audits, performance analysis, and JavaScript cryptography analysis. It includes a built-in MCP Server intended to expose capture and analysis capabilities to clients such as Claude Desktop and Cursor, but the source does not provide tool names or a server configuration example.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 6ae7becb2564 Data synced 2026-10-19 Read the FMRS scoring method