Skip to main content
Glama
joemunene-by

ghostsoc-mcp

by joemunene-by
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Hybrid

ghostsoc

Autonomous security analyst that drives the ghostsuite tools, correlates their findings across tools, and writes one ranked report.

Every ghostsuite tool finds something and emits it into the void. ghostsoc is the layer that thinks across them. It runs the right scanners for a target, normalises every tool's output into one findings model, chains findings that share an asset into attack paths, lifts the severity of chains that genuinely escalate, and writes a single report a human or an agent can act on.

How it works

adapters        normalise       correlate                 reason        report
ghostrecon  ->  Finding    ->   dedupe by fingerprint  -> triage per -> md / json
ghostdlp        objects         chain by shared asset     chain          / html
ghostmap                        lift escalating chains    (GhostLM or
(any SARIF tool)                                           fallback)

The one idea that makes this more than a wrapper: correlation. A leaked credential is a High on its own. A leaked credential on a host a recon pass already flagged as a reachable service is a different animal, the key unlocks something you can get to. When findings on one asset span two of exposure, secret and vuln, ghostsoc raises the chain a severity band and says why.

Related MCP server: EVIDIQ Sentinel MCP

Install

pip install -e .          # core, no third-party deps
pip install -e .[mcp]     # add the MCP server
pip install -e .[dev]     # tests and lint

ghostsoc shells out to the ghostsuite tools. It only runs the ones it finds on your PATH and quietly skips the rest, so a partial install still produces a report.

Use

ghostsoc adapters                      # what is wired in and what is installed
ghostsoc scan api.example.com          # markdown report to stdout
ghostsoc scan ./repo --format json     # json for a pipeline
ghostsoc scan api.example.com --format html --out report.html
ghostsoc scan api.example.com --only ghostrecon,ghostdlp

Adapters

adapter

tool

input

categories

output

ghostrecon

ghostrecon

host

recon, attack-surface

json

ghostdlp

ghostdlp

path, host

data-leak, secrets

json

ghostmap

ghostmap

url

web, dast

sarif

any

any

varies

sast, dast, ...

sarif

Any suite tool that emits SARIF 2.1.0 plugs in as a four-line SarifAdapter subclass. More first-class adapters (ghostsbom, ghostcloud, ghostchain) are next.

A worked example

Two tools, two findings, one host:

ghostrecon: open service on api.example.com:8080   (MEDIUM, attack-surface)
ghostdlp:   leaked database credential             (HIGH,   data-leak)

Report:

## 1. api.example.com

**CRITICAL** · ghostdlp, ghostrecon

ghostrecon found exposed service and ghostdlp found leaked database credential on the same asset.

> A leaked secret sits on an asset that is reachable from outside. Treat it as an
> active foothold: rotate the credential and close the exposure before anything else.

- **HIGH** ghostdlp: leaked database credential
- **MEDIUM** ghostrecon: exposed service

Neither tool would have told you this on its own.

Drive it from an agent

ghostsoc-mcp

Exposes scan_target(target, only=None) and list_adapters() over MCP. Point GhostLM, Claude, or any MCP client at it and the agent runs the suite and reads back correlated chains with a triage note on each.

Layout

ghostsoc/
  findings.py     the one Finding shape every adapter normalises into
  adapters/       one module per tool, plus generic SARIF ingest
  correlate.py    dedupe by fingerprint, chain by shared asset, lift escalations
  reason.py       GhostLM triage note per chain, heuristic fallback offline
  engine.py       select adapters, run in parallel, correlate, isolate failures
  report.py       markdown, json and html renderers
  cli.py          ghostsoc scan / adapters
  mcp_server.py   the MCP surface

Status

Early. Core pipeline (adapters, correlation, reasoning, reports, CLI, MCP) works. Next: more first-class adapters, a YAML policy gate reusing secure-mcp, and a coverage view once ghostrange lands. See the roadmap in the issues.

License

MIT

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • Heron by heronsec

    Heronofficial

    heronsec
    A
    license
    Not graded
    quality
    D
    maintenance
    Security scanning for AI agent skills, MCP servers, and agent prompts, returning signed trust scores and detailed findings.
    MIT
  • EVIDIQ Sentinel MCP by evidiq

    EVIDIQ Sentinel MCPofficial

    evidiq
    A
    license
    Not graded
    quality
    B
    maintenance
    Scans remote MCP endpoints, manifests, and agent skills for security threats, providing deterministic scores, verdicts, and signed reports to verify agent infrastructure before trust or payments.
    1
    MIT

  • mcp-security-server

    Balckers
    F
    license
    Not graded
    quality
    C
    maintenance
    Provides AI agents with standardized security scanning capabilities for authorized assessments, including host detection, port scanning, service fingerprinting, web vulnerability scanning, CVE verification, and report generation, all with unified JSON output and strict compliance controls.
    -

  • agent-sec-mcp

    sudo-ai-git
    A
    license
    Not graded
    quality
    C
    maintenance
    Enables authorized security auditing of AI-agent supply chains and agent-facing surfaces: deterministic local skill-bundle audits against eight attack patterns, secrets scanning, and scope-gated read-only recon of agent endpoints and MCP surfaces.
    MIT