tenrec 1.0.1

A modern MCP server implementation for IDA Pro using ida-domain and FastMCP.

Tenrec

A headless, extendable, multi-session, IDA Pro MCP framework built with ida-domain and FastMCP, inspired by mrexodia/ida-pro-mcp. Supports custom plugins for easy extension.

Table of Contents

• Overview
• Demo
• Installation
• Usage
  • install / uninstall
  • run
  • docs
  • plugins
• Creating Custom Plugins
• Testing
• Known Issues
• Contributing
• Future Work

Overview

Tenrec aims to simplify the MCP experience with IDA Pro. Inspired by the existing IDA Pro plugin ecosystem and ida-pro-mcp, Tenrec provides a robust framework for building and using plugins to automate and enhance reverse engineering tasks with LLMs.

Key features:

• Completely headless IDA Pro interaction, made possible with ida-domain
• Multi-session analysis, great when working with libraries used by an application
• Custom plugin support
• Auto-docs generation (including plugin docs)

Out-of-the box, Tenrec includes the following core plugins:

• Functions: Analyze and manage functions, including boundaries, attributes, and pseudocode generation.
• Cross-References (Xrefs): Trace data and code cross-references throughout the binary.
• Names: Manage symbol names, demangling, and naming conventions.
• Comments: Add and retrieve various types of comments for documentation.
• Strings: Search and analyze string literals in the binary.
• Segments: Query memory segments and sections.
• Bytes: Read and patch binary data at any address.
• Types: Manage type information and data structures.
• Entries: Access entry points and exported functions.

For a complete breakdown of available plugins and their operations, check out the documentation.

[!NOTE]
For a list of community plugins, see the tenrec-plugins repository.

Demo

Using tenrec and Claude Code, we were able to solve Challenge 4 from Flare-On 9 (darn mice) with a single prompt (see below). Find the complete challenge write-up here.

To test this yourself, checkout the fareedfauzi/Flare-On-Challenges repo for the binary and additional resources.

You have been provided with the binary called darn_mice.exe in the current directory. Use tenrec to open a session and reverse the binary. Focus on getting a high-level understanding of the application by finding entry points, and tracing execution flow. Make sure to rename variables and functions as you work through the binary. The xref plugin can be helpful in this task. You will ultimately be looking for a flag. The flag is formatted in an email format, and may require decryption. Do not try to guess the flag, use your analysis to guide you towards the correct answer based on the binary.

https://github.com/user-attachments/assets/3eb442dd-9b7a-44a6-836b-b73f99f4c2f3

Installation

Prerequisites

• Python 3.10 or higher
• IDA Pro 9.1+ installation

Install with uv

uv tool install tenrec

Confirm that tenrec is installed:

# You should see tenrec in the list of installed tools
uv tool list 
# Which tenrec should show the path to the tenrec executable
which tenrec

Since tenrec depends on ida-domain, you need to set the IDADIR environment variable to point to your IDA Pro installation directory before running tenrec commands. For example:

# macOS example
export IDADIR="/Applications/IDA Professional 9.1.app/Contents/MacOS"
# Linux example  
export IDADIR="/opt/idapro"  
# Windows example (PowerShell, run as Administrator)
setx IDADIR "C:\Program Files\IDA Pro" /M  

Development Installation

# Clone the repository
git clone https://github.com/axelmierczuk/tenrec.git
cd tenrec

# Install in development mode with dependencies
uv pip install -e ".[dev]"

Usage

install / uninstall

After installing tenrec, you can install or uninstall the MCP server with a variety of MCP clients.

Currently, the following clients are supported:

• CLine
• Roo Code
• Kilo Code
• Claude
• Cursor
• Windsurf
• Claude Code
• LM Studio

Additional clients like Cortex can be used as well, but require manual installation at the moment.

Get started with the install command:

tenrec install

Install / uninstall command help menus

 Usage: tenrec install [OPTIONS]

 Install tenrec with MCP clients.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ --help      Show this message and exit.                                                  │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

 Usage: tenrec uninstall [OPTIONS]

 Uninstall tenrec and MCP clients.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ --help      Show this message and exit.                                                  │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

Manual installation for Codex

tenrec can also be used with other LLM clients like Codex, even if these are not yet supported by the install command.

The following step by step installation guide shows how to use Codex with tenrec and the GPT4.1 LLM via AzureAI and API key authentication on Linux. Azure OpenAI in Codex. Please note that AzureAI API keys are different from OpenAI API keys. Ensure that you deployed your AzureID resource in one of the supported regions for Responses API

In order to use tenrec with Codex proceed as follows:

• Download the latest Codex release from https://github.com/openai/codex/releases (We're using 0.39.0 in the Linux x86_64-unknown-gnu.tar.gz version)

• Unpack at the place of your choice with: tar --zstd -xvf codex-x86_64-unknown-linux-gnu.tar.gz

• Make it executable via: chmod +x codex-x86_64-unknown-lunux

• Deploy the GPT-4.1 model in your Azure OpenAI workspace:

  See https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/create-resource?pivots=web-portal for details. Make sure you are using a supported region for the Responses API

• Create the codex configuration file in ~/.codex/config.toml with the following content

  # Set the default model and provider
  model = "gpt-4.1" 
  model_provider = "azure"
  preferred_auth_method = "apikey"
  
  # Configure the Azure provider
  [model_providers.azure]
  name = "Azure"
  # Make sure you set the appropriate subdomain for this URL.
  base_url = "https://<your_azure_ai_deployment>.openai.azure.com/openai"
  env_key = "AZURE_OPENAI_API_KEY"
  query_params = { api-version = "2025-04-01-preview" }
  wire_api = "responses"
  model_reasoning_effort = "high"
  
  [projects."/tmp"]
  trust_level = "trusted"
  
  [mcp_servers.tenrec]
  command = "uvx"
  args = ["tenrec", "run"]
  env = { "IDADIR"="/home/user/ida-essential-9.2" }
  
  
  

• Set the environment variable AZURE_OPENAI_API_KEY

  You will find the API Key in the Azure AIFoundry Overview page for your deployed resource.

  export AZURE_OPENAI_API_KEY=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

  The exported variable must match the name for env_key in the config.toml file.

  Keep this API key secret.

• Test it

  codex-x86_64-unknown-linux "Write me a Haiku"

  
  ╭────────────────────────────────────────────────────────╮
  │ >_ OpenAI Codex (v0.39.0)                              │
  │                                                        │
  │ model:     gpt-4.1   /model to change                  │
  │ directory: /tmp                                        │
  ╰────────────────────────────────────────────────────────╯
  
    To get started, describe a task or try one of these commands:
  
    /init - create an AGENTS.md file with instructions for Codex
    /status - show current session configuration
    /approvals - choose what Codex can do without approval
    /model - choose what model and reasoning effort to use
  
  ▌ Write me a Haiku
  
  > Gentle morning light
    Whispers through the waking trees
    Hope blooms with the dawn
  
  ▌ Find and fix a bug in @filename                                                                                                                                            
  
  ⏎ send   Ctrl+J newline   Ctrl+T transcript   Ctrl+C quit   14.0K tokens used   99% context left                                                                             
  

  If you receive a response without error, Codex has been setup correctly with AzureAI.

• Test tenrec in Codex

  Use codex with the same demo prompt above. Copy darn_mice.exe to /tmp and run

  codex "You have been provided with the binary called darn_mice.exe in the /tmp directory. Use tenrec to open a session and reverse the binary. Focus on getting a high-level understanding of the application by finding entry points, and tracing execution flow. Make sure to rename variables and functions as you work through the binary. The xref plugin can be helpful in this task. You will ultimately be looking for a flag. The flag is formatted in an email format, and may require decryption. Do not try to guess the flag, use your analysis to guide you towards the correct answer based on the binary."

run

Run the tenrec MCP server!

By default, tenrec loads built-in plugins and ones stored in the config. If you want to disable loading the config or built-in plugins, use the --no-config or --no-default-plugins flags. Additional plugins can be specified with the -p / --plugin argument.

tenrec run --transport sse

Run command help menu

 Usage: tenrec run [OPTIONS]

 Run the tenrec server.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ --no-config                                               If set, the configuration file │
│                                                           will not be used to load       │
│                                                           plugins.                       │
│ --no-default-plugins                                      If set, default plugins will   │
│                                                           not be loaded.                 │
│ --transport           -t  [stdio|http|sse|streamable-htt  Transport type to use for      │
│                           p]                              communication (default: stdio) │
│ --plugin              -p  TEXT                            Plugin to load. Could be a     │
│                                                           PyPI package name, local path, │
│                                                           or git repo.                   │
│ --help                                                    Show this message and exit.    │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

plugins

Tenrec supports a plugin system that allows you to extend its functionality with custom plugins.

Adding plugins is made simple by specifying a package name that uv can process. For example, this can be a package that can be found on pypi, a git repo, or on the filesystem. Under the hood, tenrec uses uv to install plugins, so you can use any format supported by uv (Astral - Managing Packages).

Examples

tenrec plugins add --plugin "example-package"       # Install from PyPI
tenrec plugins add --plugin "/path/to/local/plugin" # Install from local path
tenrec plugins add --plugin \                       # Install from git repo
  "git+ssh://[e-mail]/axelmierczuk/tenrec#subdirectory=examples"

# Install from git repo
tenrec plugins add --plugin \
  "git+ssh://[e-mail]/axelmierczuk/tenrec#subdirectory=examples"
# List installed plugins
tenrec plugins list
# Remove a plugin by dist name
tenrec plugins remove --dist example_plugin

Plugins command help menus

 Usage: tenrec plugins [OPTIONS] COMMAND [ARGS]...

 Manage tenrec plugins.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ --help      Show this message and exit.                                                  │
╰──────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Commands ───────────────────────────────────────────────────────────────────────────────╮
│ add                 Add a new plugin.                                                    │
│ list                List installed plugins.                                              │
│ remove              Remove an existing plugin.                                           │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

 Usage: tenrec plugins add [OPTIONS]

 Add a new plugin.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ *  --plugin  -p  TEXT  Plugin to load. Could be a PyPI package name, local path, or git  │
│                        repo.                                                             │
│                        [required]                                                        │
│    --help              Show this message and exit.                                       │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

 Usage: tenrec plugins list [OPTIONS]

 List installed plugins.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ --help      Show this message and exit.                                                  │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

 Usage: tenrec plugins remove [OPTIONS]

 Remove an existing plugin.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│ *  --dist  -d  TEXT  Plugin dists(s) to remove from the configuration [required]         │
│    --help            Show this message and exit.                                         │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

docs

The docs command generates documentation for your plugins and operations using docsify. This allows you to create self-documenting plugins that are easy to understand and use.

Using the -p / --plugin argument, you can specify one or more plugins to include in the documentation. Since the server implements session management as core functionality, it will automatically be included in your documentation.

Not happy with the default homepage? You can specify a custom README file with the -r / --readme argument.

git clone [e-mail]:axelmierczuk/tenrec.git
cd tenrec
tenrec docs -p tenrec/plugins/plugins

Docs command help menu

 Usage: tenrec docs [OPTIONS]

 Generate documentation.

╭─ Options ────────────────────────────────────────────────────────────────────────────────╮
│    --output     -o  DIRECTORY  Output directory for the generated documentation.         │
│                                [default: docs]                                           │
│    --readme         PATH       Path to a README file to include in the documentation.    │
│    --base-path      TEXT       The base path for the URL.                                │
│    --repo           TEXT       The URL of the repository for the project.                │
│    --name           TEXT       Name of the documentation set. [default: tenrec]          │
│ *  --plugin     -p  TEXT       Plugin to load. Could be a PyPI package name, local path, │
│                                or git repo.                                              │
│                                [required]                                                │
│    --help                      Show this message and exit.                               │
╰──────────────────────────────────────────────────────────────────────────────────────────╯

Creating Custom Plugins

Extend Tenrec with your own plugins!

Simply build a class inheriting from PluginBase and define operations.

There are a few important things to note:

• Name
  • The name attribute must be in snake_case format. This name will be pre-pended to all operation names to avoid conflicts.
• Description
  • The class docstring will be used as the plugin description in the documentation.
• Instructions
  • The instructions attribute helps LLMs understand the purpose and usage of your plugin.
• Operations
  • Each operation must be decorated with the @operation() decorator. You can have other methods in your class, but only decorated methods will be exposed as operations.
  • The docstrings of the operations will be used by the model to understand what the operation does. Be accurate with your parameters and return types.
• Parameters / Return Types

  • When working with addresses, use the HexEA model if possible. This class ensures that addresses are always represented in hexadecimal format, which is more familiar to reverse engineers and the LLMs. It also integrates directly with the ida-domain database APIs. Import it with:

    from tenrec.plugins.models import HexEA
    

  • Use standard Python types (e.g., int, str, list, dict) or Pydantic models for parameters and return types.

Once your plugin is defined, create a pyproject.toml file to package it, making sure to include:

[project.entry-points."tenrec.plugins"]
plugin = "file:ClassName" # Specify the path to your plugin class

A more complex example can be found in examples.

Example plugin

from tenrec.plugins.models import PluginBase, Instructions, operation

class CustomAnalysisPlugin(PluginBase):
    """This docstring will be used as the description of your plugin when documenting!"""
    
    name = "must_be_snake_case"
    version = "1.0.0"
    
    # Instructions help LLMs understand the purpose and usage of your plugin
    instructions = Instructions(
        purpose="Perform custom binary analysis operations",
        interaction_style=[
            "You can call operations like find_crypto() or detect_packer() to analyze the binary."
        ],
        examples=[
            "Find cryptographic constants: custom_analysis_find_crypto()",
            "Detect packers: custom_analysis_detect_packer()"
        ],
        anti_examples=[
            "Some anti examples",
        ]
    )
    
    @operation()
    def find_crypto(self) -> list[dict]:
        """Find potential cryptographic constants in the binary."""
        results = []
        # Your analysis logic here using self.database, which is provided to you by PluginBase
        # When switching to a different database/session, this will be updated automatically
        for ea in self.database.functions.get_all():
            # Check for crypto patterns
            pass
        return results
    
    @operation()
    def detect_packer(self) -> dict:
        """Detect if the binary is packed."""
        # Packer detection logic
        return {"packed": False, "packer": None}

Creating Custom Operation Parameters

If you find yourself needing to add the same arguments to many of your operations, you can define custom operation parameters (see tenrec/plugins/models/parameters.py for the implementation details of PaginatedParameter).

Custom operations must implement four methods:

• hook_apply_signature
• hook_apply_annotations
• hook_pre_call
• hook_post_call

hook_apply_signature

hook_apply_signature(self, signature: inspect.Signature) -> inspect.Signature

The hook_apply_signature hook allows for modification of the operation signature. You can add or remove parameters, change their types, or modify the return type. In the PaginatedParameter, we use this hook to add offset and limit parameters to the operation signature.

hook_apply_annotations

hook_apply_annotations(self, annotations: dict) -> dict

The hook_apply_annotations hook allows for modification of the operation annotations. In the PaginatedParameter, we use this hook to update the return type annotation to reflect the pagination.

hook_pre_call

hook_pre_call(self, context: dict, *args, **kwargs) -> tuple[tuple, dict]

The hook_pre_call hook allows for modification of the operation arguments before the call to the operation. In the PaginatedParameter, we use this hook to pop the offset and limit parameters from the arguments into the context that is passed to the operation.

hook_post_call

hook_post_call(self, context: dict, result: Any) -> Any

The hook_post_call hook allows for modification of the operation result after the call to the operation. In the PaginatedParameter, we use this hook to slice the result based on the offset and limit values from the context.

Testing

Tenrec includes a comprehensive test suite. Run tests with pytest:

# Run all tests
IDADIR="/path/to/ida" uv run pytest tenrec/tests/

# Run specific test module
IDADIR="/path/to/ida" uv run pytest tenrec/tests/unit/test_functions_plugin.py

# Run with coverage report
IDADIR="/path/to/ida" uv run pytest tenrec/tests/ --cov=tenrec --cov-report=html

# Run tests with verbose output
IDADIR="/path/to/ida" uv run pytest tenrec/tests/ -v

Writing Tests

Create unit tests for your custom plugins:

import pytest
from unittest.mock import MagicMock
from my_plugins import CustomAnalysisPlugin

class TestCustomAnalysisPlugin:
    @pytest.fixture
    def plugin(self):
        plugin = CustomAnalysisPlugin()
        plugin.db = MagicMock()
        return plugin
    
    def test_find_crypto(self, plugin):
        # Mock IDA database calls
        plugin.db.functions.get_all.return_value = [0x401000, 0x402000]
        
        # Test the operation
        results = plugin.find_crypto()
        assert isinstance(results, list)

Known Issues

• Tenrec is currently not supported on Windows. Work to enable this is ongoing. See Issue #9 for details.

Contributing

We welcome contributions! Please follow these guidelines:

Development Setup

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Make your changes
4. Run tests and linting
5. Commit your changes (git commit -m 'Add amazing feature')
6. Push to the branch (git push origin feature/amazing-feature)
7. Open a Pull Request

Code Style

We use ruff for linting and formatting:

# Check code style
ruff check .

# Format code
ruff format .

# Fix common issues
ruff check --fix .

Commit Guidelines

• Use clear, descriptive commit messages
• Reference issues when applicable
• Keep commits focused and atomic

Future Work

• Enhanced Plugin System: Dynamic plugin loading and hot-reload support
• Performance Optimizations: Caching and batch operations
• Additional Plugins Ideas:
  • Vulnerability detection
  • Binary diffing
  • Automated unpacking
  • Machine learning-based analysis
• Integration Improvements: Better integration with other reverse engineering tools
• Documentation: Expanded API documentation and tutorials