
redamon
Self-hosted AI pentesting framework that graphs attack surfaces and opens fix PRs
The repository was last updated 4 days ago and is published under MIT.
- Pricing
- Open source
- Platforms
- Linux, macOS, Windows
- GitHub stars
- 3.4K
- Last commit
- 4 days ago
- Language
- Python
- License
- MIT
Similar MCP servers
About redamon
redamon is an open-source MCP server for ai models and llm tooling. It is available on Linux, macOS, Windows, self-hosted, API, CLI and Web.
The README and site position RedAmon as a self-hosted penetration testing platform run through Docker Compose. Recon tools feed findings into a Neo4j attack-surface graph, which an autonomous agent queries before choosing exploitation steps. Dangerous tools and phase changes sit behind human approval gates, and Rules of Engagement limits are enforced in code.
The project publishes a result on the XBOW validation benchmark, reporting 101 of 104 targets solved fully black-box, with unedited agent sessions linked for each solve. A separate CypherFix stage triages findings and has a code agent propose a fix as a GitHub pull request against the target repository.
It suits security teams and researchers who have written authorization to test their own infrastructure and want an auditable, locally hosted workflow. Local Ollama or vLLM models can be used so prompts and results stay in-house.
What it does
- Attack surface graph
- Every recon finding is stored as a typed node in Neo4j and queried by the agent in plain English or Cypher.
- Approval-gated agent
- An autonomous agent runs exploitation steps, but Metasploit, Hydra, code execution and phase changes need human sign-off.
- Rules of Engagement enforcement
- Uploaded RoE documents are parsed into scope, time window and rate limit constraints checked by the tool executor.
- Parallel recon pipeline
- Dozens of scanners run in stages, with each tool's output passed as input to the next.
- CypherFix pull requests
- A triage agent ranks findings and a CodeFix agent proposes a patch that is opened as a GitHub pull request.
- Local model support
- Works with OpenAI-compatible endpoints and local Ollama or vLLM models for air-gapped use.
Good for
- Mapping the external attack surface of an owned domain or IP range before a scheduled assessment
- Running authorized web application tests with an agent that chains recon into exploitation
- Triaging scanner output into a ranked remediation list
- Generating a pentest report from graph data for a client or internal review
- Opening pull requests that patch vulnerabilities found in your own repository
Pricing
Open source
MIT-licensed, self-hosted; no paid hosted plan mentioned.
Prices change. Check redamon.org before you buy.
redamon FAQ
Is RedAmon free and open source?
The README says it is released under the MIT license and the source is on GitHub.
Do I need a cloud LLM account to run it?
No. The project says a scan runs without external accounts and supports local models through Ollama, vLLM and LM Studio.
What does the XBOW result mean?
The site reports that RedAmon solved 101 of 104 XBOW validation targets fully black-box, with an unedited session published for each solve.
What system resources are needed?
The README lists 2 CPU cores and 7 GB RAM minimum without OpenVAS, with 80 GB free disk, and Docker Engine 28.0.0 or newer.
Is redamon free?
Yes. redamon is open source, so you can use and self-host it for free.
What are the best alternatives to redamon?
The closest alternatives to redamon are AiSOC, Redthread, prompt-protection, Redproof and CybRelay.
Where this comes from
- Description: written from the project's README and the project's website.

