shannon vs pentagi
Two of the top projects in Security, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
shannon
Autonomous pentester that exploits web apps and APIs using source code analysis
pentagi
Autonomous multi-agent penetration testing system with sandboxed security tools
Verdict
shannon ranks higher, 73 to 55. Of the 3 score parts we measure for both, shannon leads on 2, pentagi on none and 1 is even.
- Adoptionshannon leads by 46 points, 61 to 15: a wide gap.
- FreshnessEven at 100.
- MaintenanceWe do not score pentagi on this; shannon has 93.
- Easy to runWe do not score pentagi on this; shannon has 50.
- Agent-readyshannon leads by 15 points, 55 to 40: a wide gap.
Pick shannon if you are
Developers and security teams pentesting apps they own, locally or in CI
- Reports only exploits with a reproducible proof of concept, which cuts speculative scanner noise
- Outputs PDF, Markdown, JSON and SARIF 2.1.0; GitHub Action and GitLab CI can gate on exploited findings
Pick pentagi if you are
Security engineers, pentesters and red teams doing authorized testing
- Runs 20+ pentest tools (nmap, Metasploit, sqlmap) in an isolated Docker sandbox
- Supports 10+ LLM providers, custom OpenAI-compatible endpoints, and local Ollama or vLLM
| What we compare | shannon | pentagi |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 61, popular | 15, niche |
| Freshness | 100, active | 100, active |
| Maintenance | 93, healthy | Not scored for this kind |
| Easy to run | 50, easy | Not scored for this kind |
| Agent-ready | 55, partly | 40, minimal |
| Facts from GitHub and the README | ||
| Stars | 48.9k | 25.3k |
| License | AGPL-3.0 (copyleft) | MIT (permissive) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Oct 2026 | Oct 2026 |
| Language | TypeScript | Go |
| Docker | Yes | Yes |
| GPU | Not needed | Optional |
| arm64 or Apple Silicon | Not stated | Not stated |
shannon
Shannon reads a web application's source, maps attack paths, then uses browser automation and command-line tools to run real exploits against the running app and its APIs. Only findings with a working proof of concept reach the PDF, Markdown, JSON or SARIF 2.1.0 report. It runs from an npx launcher with an ephemeral Docker worker, and also ships a GitHub Action and a GitLab CI component.
Who it is for: Developers and security teams pentesting apps they own, locally or in CI
Strengths
- Reports only exploits with a reproducible proof of concept, which cuts speculative scanner noise
- Outputs PDF, Markdown, JSON and SARIF 2.1.0; GitHub Action and GitLab CI can gate on exploited findings
- Bring your own key: Anthropic, OpenAI, xAI, Bedrock, gateways, or local Ollama, vLLM and LM Studio
- Supports authenticated scans (login flows, TOTP) and resumable workspaces
Weaknesses
- Needs source code of the target; black-box testing is only in paid Pro and Enterprise
- Actively executes exploits, so it must not be pointed at production systems
- Anthropic and OpenAI cyber safeguards can interrupt scans unless you complete their verification
- Claude Code subscriptions unsupported after version 1.9.0; model API costs are yours
- no GPU
- Docker + Compose
- Needs Docker, Node.js 18+, AI provider API key or subscription
- Models: Anthropic, OpenAI, xAI, AWS Bedrock, Ollama
pentagi
PentAGI takes a target and goal in plain language, then uses a team of specialized agents (orchestrator, researcher, developer, executor) to plan and run 20+ tools such as nmap, Metasploit and sqlmap inside an isolated Docker sandbox. Findings are stored in PostgreSQL with pgvector, with an optional Neo4j/Graphiti knowledge graph, and it produces vulnerability reports. It works with 10+ LLM providers, any OpenAI-compatible endpoint, or local models via Ollama or vLLM.
Who it is for: Security engineers, pentesters and red teams doing authorized testing
Strengths
- Runs 20+ pentest tools (nmap, Metasploit, sqlmap) in an isolated Docker sandbox
- Supports 10+ LLM providers, custom OpenAI-compatible endpoints, and local Ollama or vLLM
- Persistent memory in PostgreSQL with pgvector, optional Neo4j knowledge graph via Graphiti
- REST and GraphQL APIs with Bearer tokens; Langfuse, Grafana and Prometheus integrations
Weaknesses
- Full stack spans many containers (Postgres, Neo4j, Langfuse, ClickHouse, Grafana), heavy to self-host
- Agent profiles, plugins, MCP servers and attack-chain view are Enterprise-only
- Not a BAS or adversary emulation tool; no predefined campaigns
- Report export limited to web, Markdown and PDF; no JSON export
- GPU optional
- Docker + Compose
- Needs Docker, Docker Compose, PostgreSQL with pgvector, Neo4j (optional), Langfuse (optional), Grafana/Prometheus (optional)
- Models: OpenAI, Anthropic, Google Gemini, AWS Bedrock, Ollama
More in Security
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.