frida-mcp

Frida MCP

Overview

Use frida-mcp when an AI client needs direct access to Frida process, device, and scripting operations. Prefer a dedicated virtual environment under ~/.local/opt/frida-mcp so the command path stays stable across client registrations.

Quick Workflow

1. Confirm core Frida tooling works first (frida-ps or frida --version).
2. Install or update the MCP package with tools/mcp/reverse-engineering-mcp-bundle/scripts/install_python_mcp.py.
3. Register manifests/frida-mcp.json into Codex, Claude, and Gemini with the shared registration helper.
4. Verify the client can at least enumerate devices or processes before using it for deep instrumentation.
5. Switch to the plain frida skill when the work is about scripts, hooks, or runtime technique rather than MCP setup.

Common Mistakes

• Installing frida or frida-tools and assuming the MCP layer is included.
• Forgetting to confirm transport context (local, USB, remote) before judging the MCP server.
• Treating package install as complete without registering the client entry.

Reference Map

Read references/upstream.md for the upstream install snippet and the repo bundle commands.