Skip to main content
Glama
ReversePlumbing
by chibitek

ReversePlumbing is a local MCP server: it gives an AI client structured information about files you allow it to read. Written from scratch by Chibitek, it helps turn evidence into an original product specification. It does not reproduce the inspected product or prove how its server works.

Quickstart

Pre-release source checkout. Requires Node.js 22.14 or newer. An npm release is not available through this README yet. Start inside a reviewed checkout of an approved commit:

npm ci --ignore-scripts --no-audit --no-fund
npm run build
node dist/server.js --root /absolute/path/to/isolated-artifacts

Replace the artifact path with an existing folder you intend to inspect. The process speaks MCP over standard input/output; it is not an interactive terminal prompt or a web server. Client setup and network-denied execution. Installation downloads dependencies; the running server itself makes no network requests.

Related MCP server: REA

What you can do

Job

Result

Limit

Inspect Apple apps

Bundle identity, executable structure and signing metadata

Metadata does not prove a user workflow or certificate trust

Inspect Windows software

PE structure, imports, .NET metadata and signature digest comparison

No execution or certificate-chain validation

Inspect Linux libraries and programs

ELF structure, dependencies, symbols and hardening flags

No execution

Inspect Android packages

Manifest, DEX inventory, native libraries and signing structures

No decompilation or signing trust verification

Inspect Electron and JavaScript packaging

Archive inventory and locally available source maps

Remote source maps are never downloaded

Map saved browser traffic

Templated endpoints, page call sequences, JSON shapes and observed auth schemes

Values are omitted; unknown names are generalized; capture order does not prove causality

Review visual evidence

Reviewed PNGs returned as images, accessibility snapshots and hashed source files

Capture and redaction happen separately; no automatic OCR

Map tested workflows

Before/after evidence, actions, feature claims and explicit coverage gaps

Operator-authored claims; only declared workflows count

Opt into deeper native analysis

Functions identified by an existing Ghidra installation

macOS only; explicit opt-in and network-denied sandbox required

How it works

flowchart LR
  A[Files in approved folders] --> B[Read-only parsers]
  B --> C[Structured evidence]
  C --> D[Your MCP client]
  D --> E[Your model provider, depending on client]

The server exposes only stdio transport. It does not connect to a browser, fetch packages at runtime, or execute an inspected app. Ghidra is the single optional subprocess and must be installed separately.

Choose the right approach

Need

Approach

Understand package structure or dependencies

Start with the built-in readers

Identify native functions

Explicitly opt into the existing Ghidra installation

Verify an actual user workflow

Capture the licensed app separately, then import workflow evidence

Design an original replacement

Use the clean-room method and an evidence-backed behavior spec

Privacy and limits

Approve isolated folders only, never a home directory or a folder holding credentials. Reads are bounded and checked against approved roots. This is not a complete filesystem sandbox against another process changing ancestor directories during a read. Use isolated, stable copies for untrusted inputs.

Output goes to your AI client and may reach its model provider. Get permission before inspecting a confidential target. Use only software and services you own, license or are authorized to examine, within their terms. Never copy their code, wording, assets or paid content, or bypass access and license controls. Security policy.

FAQ

Does it change the app or call its servers? No. Inspection is read-only and the runtime performs no outbound networking. The optional analysis subprocess is subject to network denial.

Does it prove an app is safe? No. Signing fields are structural evidence, not a trust verdict. Read the per-format limits.

Does it run everywhere? Built-in readers work on local files. Verification has covered macOS and an isolated Linux container; do not infer full native Windows support from PE parsing. The optional Ghidra invocation requires the macOS sandbox.

How do I check my checkout? Run npm test. Optional tests depend on installed tools; review pass and skip counts. No hosted build badge is used.

Contributing and security

See Contributing and Security. Detailed parser coverage, verification commands and operational limits are in the technical reference.

License

Apache License 2.0. Built by Chibitek Labs, the product and innovation arm of Chibitek.

Tool Schema Changelog

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

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI clients to inspect local software structure, code, and runtime behavior through evidence-backed analysis, including sources, coverage, and limitations.
      1
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Connects AI agents and CLI users to local analysis tooling that inspects native binaries, JavaScript/Electron apps, .NET assemblies, websites, and saved network captures, returning evidence-backed findings with their limitations. This lets users understand how an existing app feature works and rebuild a similar version in their own project.
      76 npm
      5
      MIT
    • A
      license
      Not graded
      quality
      Not graded
      maintenance
      Enables AI agents to inspect and reverse engineer local software targets, including binaries, JavaScript/Electron apps, .NET assemblies, and websites, returning evidence-backed findings for understanding and building similar features.
      377,174 npm
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables MCP-compatible AI assistants to inspect local application internals by analyzing JavaScript/Electron/ASAR and .NET artifacts, tracing dependencies, interfaces, and data flows, and preserving traceable evidence.
      14
      MIT