Hey Product Hunt! 👋
I'm Brooke, founder of AFaaS. Last year my security team blocked my AI assistant from accessing my Obsidian vault: "Data leaves our infrastructure. We can't allow that."
Sound familiar? Every mainstream AI integration sends your data to someone else's servers. For regulated industries (finance, healthcare, defense, government), that's a hard no.
So I built three MCP servers that run entirely on your machine:
🔐 vault-bridge — Obsidian vault (read/write/search, frontmatter, daily notes, stateless protocol)
📁 airgap-file-bridge — Filesystem (ripgrep search, atomic writes, symlink protection, stdio/SSE/HTTP)
🐘 airgap-db-bridge — PostgreSQL (parameterized queries = SQLi prevention, read-only mode, audit logs, stateless)
All three:
pip install vault-bridge airgap-file-bridge airgap-db-bridge
Standard MCP protocol → works with Claude Desktop, Cursor, VS Code
uvx airgap-file-bridge /tmp for zero-install demo
MIT licensed, 155 tests passing, CI/CD green
Published on PyPI (test + production)
Why this exists: Your data, your rules. Zero data egress. Works offline. Works in air-gapped environments. No vendor lock-in.
Built for developers who want AI with local data, security teams blocking cloud AI, and enterprises with SOC2/HIPAA/ITAR/GDPR requirements.
Happy to answer questions about architecture, security model, or MCP protocol implementation!
Links:
GitHub: Airgap Fleet · GitHub
PyPI: vault-bridge, airgap-file-bridge, airgap-db-bridge
Docs: README in each repo
Try it: pip install vault-bridge airgap-file-bridge airgap-db-bridge
What would you connect first — your vault, files, or database? 👇
Replies
GitHub