Mantis: Portable Toolkit for Building Secure Software

[!CAUTION] USE AT YOUR OWN RISK. BE EXTREMELY CAREFUL. This suite is designed to generate and execute autonomously generated code that may be unstable or perform unexpected actions. USE THIS ONLY IN ISOLATED, RESTRICTED ENVIRONMENTS. Never run this suite on a machine with access to production systems, sensitive data, or internal networks.

[!IMPORTANT] RESPONSIBLE USE AI models are non-deterministic and can hallucinate findings or generate incorrect patches. All findings must be manually verified by a security expert before being reported. Do not mass-file unverified, AI-generated reports to open-source maintainers. A failure to automatically reproduce a vulnerability does not definitively mean it is a false positive, nor does a successful reproducer guarantee the bug is exploitable in all contexts. Use Mantis responsibly.

Mantis is a set of skills along with an ADK reference harness for building secure software in the new AI era of software development.

The default model is DeepSeek (ollama/deepseek-v4-flash:cloud). This guide gets Mantis running against Ollama Cloud so you can use hosted DeepSeek models with no GPU and no local daemon. To run entirely locally instead, see Getting Started (Local).

• Python 3.10+ and python3-venv
• An Ollama Cloud account and API key from https://ollama.com/settings/keys
• (Recommended) the ollama CLI, to test connectivity: ollama login

cd reference && ./install.sh

Mantis loads credentials from a git-ignored .env file. Copy the template and fill in your key:

cp reference/.env.example reference/.env
# edit reference/.env
#   OLLAMA_API_KEY=your-ollama-cloud-key

Existing shell environment variables always win over .env, so you can also run export OLLAMA_API_KEY=your-key instead.

The default model ollama/deepseek-v4-flash:cloud is auto-routed to https://ollama.com (the Ollama-native :cloud endpoint). Confirm with a live probe after configuration (step 5) — the preflight --probe flag will verify credentials and reachability.

cd reference && source .venv/bin/activate

# Fast config & capability auto-detection (uses deepseek-v4-flash:cloud by default)
python3 scripts/configure.py --auto

# Preflight validation + live reachability probe (verifies your API key)
python3 scripts/configure.py --test --probe

If the probe reports Unauthorized, double-check the OLLAMA_API_KEY in your .env / shell and rerun.

# From reference/
./run.sh /path/to/target           # a source file or a repository directory

# Optional: target a specific objective
./run.sh /path/to/target --objective "Audit for Server-Side Request Forgery and SSRF in webhook handlers"

To use a different DeepSeek model (e.g. deepseek-v4-pro), or any other model, set it on the command line or in workflow.json:

./run.sh . --model ollama/deepseek-v4-pro:cloud

Routing refresher: ollama/<m>:cloud → native Ollama Cloud endpoint https://ollama.com; ollama.cloud/<m> → OpenAI-compatible endpoint https://ollama.com/v1; a plain ollama/<m> (no :cloud) → your local Ollama daemon (http://localhost:11434/v1), which needs a downloaded model.

Prefer an entirely local setup with no cloud account? First, install python3-venv such as with sudo apt install python3-venv, then run the install script. Mantis comes with automated configuration and launcher tools (mantis-configure and mantis-launch):

cd reference && ./install.sh

# Start a local Ollama daemon and pull a real, downloadable DeepSeek model.
# (Local Ollama needs no cloud credentials.)
ollama serve &
ollama pull deepseek-v4-flash      # local (non-cloud) build of the default model

# Fast Configuration & Capability Auto-Detection
python3 scripts/configure.py --auto --model ollama/deepseek-v4-flash

# Fast Preflight Validation + Live Reachability Probe
python3 scripts/configure.py --test --probe

# Launch Vulnerability Review Campaign (file or repository)
./run.sh path/to/code              # a file or a directory

# (Optional) Research Graph Synthesis for a Specific Objective
./run.sh path/to/code --objective "Audit for Server-Side Request Forgery and SSRF in webhook handlers"

Mantis is roughly designed to:

• Review history of a codebase to look for historical vulnerabilities we do not wish to repeat
• Build a semantic index (and/or summaries) of the codebase for efficient code navigation
• Automatically generate a threat model
• Build up a set of hypotheses for individual agents to research (or simply "scan every file")
• Execute on those research plans
• Deduplicate existing findings
• Triage/critique findings to combat hallucination and statically verify production viability
• Reproduce the vulnerability to varying degrees, depending on available environments, anywhere from a static guess at what an exploit might look up up to a unit test or even spinning up a mock server to attempt to exploit
• Look across known vulnerabilities to attempt to build more impactful chains
• Patch discovered vulnerabilities, using an adversarial loop to verify the vulnerability is really fixed
• Calibrate all findings based on an established rubric to combat LLM inflation of severity (surfacing the most critical risks to humans instead of spewing thousands of "criticals")
• Reflect on each iteration of the loop to look for things we've learned in the now completed round of research
• Based on all of the collected learnings, threat model, and knowledge base, use the /mantis-advise skill to develop code more securely and ensure that during development you do not repeat prior mistakes or trigger edge cases in code that were previously protected by some guard that was removed

Additionally, the ADK reference harness shows some of the neat ways in which we can build agentic workflows. Specifically, reference/workflow.json shows how a deep review is done, and research graph synthesis (./run.sh target --objective "...") allows generating custom graph topologies on the fly. This is a very powerful construct because you can use this to build any kind of agentic deep dive security review you might imagine.

Mantis is intended to be a starting point rather than a rigid set of instructions. You should adapt, tune, and extend this harness to fit your organization's specific software or hardware stack. We provide an ADK-based reference harness which works out of the box, but any competent coding agent should be able to convert this to use your framework of choice.

The Mantis skills can be adapted to specialized domains (such as Hardware/RTL, Infrastructure as Code, ML pipelines, or compiled firmware).

We strongly recommend using AI to iterate on these skills and using your internal documentation, coding standards, and build systems to augment the threat models you use for scanning. We also strongly recommend adapting risk calibration to your environment and risk tolerance.

Above all, while orchestrated vulnerability discovery is incredibly powerful and useful, it is even more important to use this in a suitably isolated environment to prevent impacting production systems. The ADK reference harness provides some example sandboxes for testing, but if you use the most advanced frontier models you should go beyond this and additionally set up an additional sandboxing layer that itself contains very strong monitoring to look for escape attempts.

This is not an officially supported Google product. This project is not eligible for the Google Open Source Software Vulnerability Rewards Program.

This project is intended for demonstration purposes only. It is not intended for use in a production environment.