[Enhancement]: MCP Client Support — Connect PentAGI to External Security Tools via Model Context Protocol

[Mosstrow]

Target Component

Core Services (Frontend UI/Backend API)

Enhancement Description

Summary

Add a generic MCP (Model Context Protocol) client to PentAGI so that agents can
use any MCP-compatible external tool as a first-class tool — starting with
Burp Suite Pro as a concrete use case.

Motivation

PentAGI's current toolset covers most of the pentest lifecycle well, but there
is a growing ecosystem of professional security tools that are exposing MCP
servers as their AI integration interface. The most notable example is
Burp Suite Pro, which PortSwigger has officially published as an MCP server
extension:
https://github.com/PortSwigger/mcp-server

Without MCP client support, integrating Burp Suite (or any other MCP-compatible
tool) requires either:

• Manual CLI scripting generated by the coder agent (fragile, high overhead)
• Forking PentAGI to add vendor-specific tool implementations

With MCP client support, PentAGI gains access to the entire MCP ecosystem
through a single integration point.

Proposed Solution

Add a configurable MCP client layer to PentAGI that:

1. Reads MCP server connections from config (similar to how LLM providers
   are configured in .env / custom.provider.yml)
2. Auto-discovers available tools from each connected MCP server at startup
   via the MCP tools/list method
3. Exposes discovered tools to all agents as first-class callable tools
   alongside native tools (terminal, file, browser, delegation)
4. Supports standard MCP transports: stdio, SSE, and HTTP

Example config (.env):
MCP_SERVERS=burp,nuclei
MCP_BURP_TRANSPORT=http
MCP_BURP_URL=http://host.docker.internal:1337
MCP_NUCLEI_TRANSPORT=stdio
MCP_NUCLEI_CMD=nuclei-mcp-server

Concrete Use Case: Burp Suite Pro

With Burp Suite's MCP server running on the analyst's machine and PentAGI
connected to it, the pentester agent could:

• Trigger active scans on URLs discovered during recon — Burp's scanner is
  significantly more capable than CLI alternatives for complex web applications
• Retrieve structured findings (vulnerability type, severity, HTTP evidence)
  and incorporate them into the engagement context
• Use Burp Collaborator for out-of-band detection — blind SSRF, blind SQLi,
  XXE — a class of vulnerabilities currently undetectable with PentAGI's
  CLI-only toolset
• Pull the crawled sitemap to seed directory enumeration with real
  discovered paths

This addresses a genuine gap: PentAGI cannot currently detect OOB/blind
vulnerabilities at all, and Burp's web crawler handles JS-heavy applications
far better than gobuster/ffuf.

Broader Impact

MCP is becoming the standard integration interface for AI-connected tooling.
Beyond Burp Suite, supporting MCP opens PentAGI to:

Tool	Capability unlocked
Burp Suite Pro	Web scanning, Collaborator OOB, sitemap
Nuclei	Template-based vulnerability detection
Shodan / Censys	Passive OSINT enrichment
Ghostwriter	Direct finding logging during engagement
Internal / custom MCP servers	Organisation-specific tooling

Rather than maintaining one-off integrations per tool, MCP client support makes
PentAGI's capabilities open-ended as the ecosystem grows.

Implementation Notes

• Go MCP client libraries exist:
  github.com/mark3labs/mcp-go is a
  mature implementation
• MCP tool descriptors (name, description, input schema) map directly onto
  PentAGI's existing tool calling structure — no architectural changes required
  to the agent system
• MCP servers running on the analyst's host are reachable from the Kali
  container via host.docker.internal (already used in the codebase for
  similar purposes)
• Security boundary: MCP tool calls would be subject to the same permission
  model as existing tools

Technical Details

No response

Designs and Mockups

No response

Alternative Solutions

No response

Verification

• I have checked that this enhancement hasn't been already proposed
• This enhancement aligns with PentAGI's goal of autonomous penetration testing
• I have considered the security implications of this enhancement
• I have provided clear use cases and benefits