What it is
MIT-licensed Python OSINT agent with REPL, CLI, Web UI, and an MCP server exposing 19 investigation tools for authorized security research.
Gabriel’s notes
Quick take: OpenOSINT is an unusually pragmatic “AI + OSINT” project: it wraps real tools/APIs behind a natural-language workflow, instead of asking you to trust vibes. I’m interested—but I’d still treat it like any other security tool that can touch the open internet and your API keys.
OpenOSINT is an OSINT agent for security researchers and analysts that exposes 19 investigation tools through a REPL, CLI, native MCP server, and a browser-based Web UI. The project emphasizes “hard-stop tool calls” where the model chooses tools, but your code executes the real tooling, making “hallucinated findings” structurally harder to produce. It supports multiple backends (Anthropic Claude by default, local Ollama, and other OpenAI-compatible endpoints). Source: project README.
I saved this under Security & privacy because OSINT automation is great… right up until it becomes accidental surveillance, key leakage, or a self-hosted foot-gun.
My current stance (honest disclosure):
- I have not personally used or independently tested OpenOSINT in day-to-day operation yet.
- I have seen claims of a “static security review” and concerns around web exposure / credentials / SSRF / Docker defaults / CI-supply-chain. Unknown / not confirmed (treat this as my cautionary note, not as a verified advisory).
What looks solid (from the docs):
- It’s explicit about authorized use only and pushes users to read the disclaimer.
- The Security Policy clarifies scope and notes it’s a BYOK tool for third-party APIs (keys stay in your environment and are passed to the upstream APIs).
- There’s a defined vuln reporting path and a coordinated disclosure process.
Good fit if you want to:
- Run “email/username/domain/IP recon” workflows without babysitting 40 tabs.
- Use an LLM to chain OSINT steps, but still execute real tools (holehe/sherlock/WHOIS, etc.).
- Expose OSINT capabilities to an MCP-capable client (e.g., Claude Desktop / Claude Code-style tooling) via a native MCP server.
- Generate repeatable outputs (reports saved after investigations) instead of one-off notes.
Pricing snapshot (auto-enriched)
The core project is free and MIT-licensed. The repo also advertises paid options like a “Complete Kit” ($55), a “Setup Sprint” ($350), and an optional commercial plan layered on top of MIT (Starter €300/year; Business €600/year; Enterprise €1,000+/year) for vendor paperwork, warranty/indemnification language, and SLA/support expectations. Separately, some integrated third-party data sources may require paid API keys (e.g., Shodan/VirusTotal/Censys), depending on what you enable.
Work-use / compliance snapshot (auto-enriched)
The project’s disclaimer frames it for legitimate security research, ethical OSINT, and education—and explicitly prohibits unauthorized surveillance/stalking, harassment, unlawful activity, and bypassing access controls or third-party terms of service. It’s provided “as is” with no warranty, and users are responsible for legal compliance. For operational security, note the Security Policy: only the latest release line is supported for security fixes, and the public demo instance is out of scope and should not receive sensitive personal data.
Alternatives (auto-enriched)
- badchars/osint-mcp-server — more “structured OSINT sources as MCP tools” (lots of discrete connectors); less of a full REPL/Web-UI productized workflow than OpenOSINT.
- LevNyyon/osint-mcp — positioned as a small, local-first MCP option; likely simpler to audit and run, but narrower than OpenOSINT’s all-in-one agent + UI approach.
Before you adopt it:
- Run it locally first (ideally in an isolated environment) and assume the Web UI should be treated like an internal tool, not a public service.
- Use dedicated, limited-blast-radius API keys for any third-party services (and rotate them if you’ve ever exposed logs/env files).
- Read the DISCLAIMER and SECURITY docs end-to-end, then decide what “authorized” means in your org (and write it down).
Sources
- https://github.com/OpenOSINT/OpenOSINT
- https://github.com/OpenOSINT/OpenOSINT/blob/main/DISCLAIMER.md
- https://github.com/OpenOSINT/OpenOSINT/blob/main/SECURITY.md
- https://github.com/OpenOSINT/OpenOSINT/blob/main/COMMERCIAL.md
- https://github.com/badchars/osint-mcp-server