Updated November 18, 2026
Glossary ToolsBest models

Ghidra · MCP server

Ghidra MCP setup with an abliterated model

Ghidra MCP, through GhidraMCP, lets Claude Code decompile and rename in the open project on a model you choose.

By AI for Reverse EngineeringUpdated November 18, 2026
On this page

What it does

The README calls it a server "for allowing LLMs to autonomously reverse engineer applications." It has 11,970 GitHub stars and an Apache-2.0 license.

Install the plugin

  1. Install Ghidra 12.1.4 and uv.
  2. Download the latest release from the GhidraMCP releases page.
  3. In Ghidra, open FileInstall Extensions, click +, and pick ghidra_12.1.4_PUBLIC_<date>_GhidraMCP.zip.
  4. Restart Ghidra, and check that GhidraMCPPlugin is enabled under FileConfigureDeveloper.
  5. Optional: change the port under EditTool OptionsGhidraMCP HTTP Server. It defaults to localhost:8080.

Connect

  1. Point your agent at an abliterated model. For Claude Code, put this in ~/.claude/settings.json and export SILK_API_KEY. Codex and OpenCode configs are on the MCP setup page.
    ~/.claude/settings.json: Kimi K3.1 Pro as the default, every model in /modelshow
    {
      "apiKeyHelper": "echo $SILK_API_KEY",
      "env": {
        "ANTHROPIC_BASE_URL": "https://api.silkcompute.com",
        "ANTHROPIC_DEFAULT_MODEL": "kimi-k3.1-pro-uncensored",
        "ANTHROPIC_DEFAULT_FABLE_MODEL": "kimi-k3.1-pro-uncensored",
        "ANTHROPIC_DEFAULT_OPUS_MODEL": "kimi-k3.1-pro-uncensored",
        "ANTHROPIC_DEFAULT_SONNET_MODEL": "kimi-k3.1-pro-uncensored",
        "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek-v4.1-flash-1017-uncensored",
        "CLAUDE_CODE_MAX_CONTEXT_TOKENS": "1048576"
      },
      "modelPicker": {
        "replaceBuiltInOptions": true,
        "options": [
          { "model": "kimi-k3.1-pro-uncensored", "label": "Kimi K3.1 Pro" },
          {
            "model": "deepseek-v4.1-flash-1017-uncensored",
            "label": "DeepSeek V4.1 Flash 1017"
          },
          {
            "model": "deepseek-v4.1-flash-1017-fast-uncensored",
            "label": "DeepSeek V4.1 Flash 1017 Fast"
          },
          { "model": "glm-5.3-uncensored", "label": "GLM-5.3" },
          { "model": "glm-5.3-flash-uncensored", "label": "GLM-5.3 Flash" },
          { "model": "hy4-uncensored", "label": "Hy4" }
        ]
      }
    }
  2. Add the MCP server.
    terminalshell
    claude mcp add ghidra -- \
      uvx --from git+https://github.com/LaurieWired/GhidraMCP \
      ghidra-mcp --ghidra-server http://127.0.0.1:8080/
  3. Open a binary in Ghidra and start claude in a working folder. /mcp should list ghidra as connected.

LaurieWired/GhidraMCP README, read November 18, 2026.

Codex and OpenCode

~/.codex/config.toml (add)toml
[mcp_servers.ghidra]
command = "uvx"
args = [
  "--from", "git+https://github.com/LaurieWired/GhidraMCP",
  "ghidra-mcp", "--ghidra-server", "http://127.0.0.1:8080/",
]
terminalshell
opencode mcp add ghidra -- \
  uvx --from git+https://github.com/LaurieWired/GhidraMCP \
  ghidra-mcp --ghidra-server http://127.0.0.1:8080/

URL and headless

For a URL client, start the server and add http://127.0.0.1:8081/mcp.

terminalshell
uvx --from git+https://github.com/LaurieWired/GhidraMCP ghidra-mcp \
  --transport streamable-http --mcp-host 127.0.0.1 --mcp-port 8081 \
  --ghidra-server http://127.0.0.1:8080/

Without the GUI, GhidraMCP analyzes a binary through PyGhidra. It needs Python 3.10 to 3.13, the extension, and GHIDRA_INSTALL_DIR.

terminalshell
uvx --python 3.13 \
  --from "ghidra-mcp[headless] @ git+https://github.com/LaurieWired/GhidraMCP" \
  ghidra-mcp --headless --binary /path/to/target

Add --project DIR to keep the project. On macOS, build the decompiler once with support/gradle/gradlew buildNatives.

Use Kimi K3.1 Pro at max effort for a whole binary. On a large file, use ReVa.