barbaros-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@barbaros-mcpWhat privacy practices does ChatGPT list in the PrivacyMatrix dataset?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
barbaros-mcp
Hosted MCP connectors for barbaros.dev, built for the Meta Muse Connector Platform.
Read-only, unauthenticated, stateless. No user data is collected or stored; each connector serves a public dataset over the Model Context Protocol (Streamable HTTP, JSON responses — no SSE, since Muse's egress proxy has been reported to hang on it).
Connectors
Connector | Endpoint | Dataset | Status |
| privacymatrix — 28 AI apps × 14 privacy questions, quote-sourced | ready to deploy |
More connectors (Unpolished AI-tell scan, WIROAM shell-portability check) are planned as separate mounts on
the same server; see uygulama_plani.md §7.
Related MCP server: Datasette MCP
Develop
npm install
npm run dev # tsx watch, http://127.0.0.1:8101
npm test # vitest
npm run build # tsc + copy static/fallback assets into dist/
npm start # node dist/server.jsEnvironment: PORT (default 8101).
Architecture
src/server.ts— Express app. One MCP endpoint per connector, mounted under its own path (/privacymatrix/mcp), plus static docs (/privacymatrix/muse.md,/privacymatrix/llms.txt,/privacymatrix/) and a shared/healthz.Each POST to an MCP endpoint creates a fresh
McpServer+StreamableHTTPServerTransport(sessionIdGenerator: undefined,enableJsonResponse: true) — stateless, no session to manage, single JSON response body.src/connectors/<name>/data.ts— loads the connector's dataset from its public GitHub repo (raw.githubusercontent.com) on startup and every 6 hours; falls back to a bundled snapshot (fallback/*.json) if the network is unavailable, so the server never starts empty.src/connectors/<name>/tools.ts— the MCP tools, registered on a per-requestMcpServer.
See connectors/muse/<name>/ for the Meta submission pack per connector (brief,
skill rules, install prompt, submission text, security note, e2e evals, icon).
Deploy (VPS)
See deploy/ for the nginx site config and pm2 process definition. Summary (full steps in
uygulama_plani.md §Adım F):
DNS:
mcp.barbaros.dev→ VPS IP (Cloudflare, DNS-only).git cloneto/var/www/barbaros-mcp,npm ci --omit=dev,npm run build.pm2 start deploy/ecosystem.config.cjs,pm2 save.nginx -twithdeploy/mcp.barbaros.dev.confinsites-enabled, then reload.certbot --nginx -d mcp.barbaros.dev.Verify with
connectors/muse/privacymatrix/EVALS.md.
License
MIT. The underlying privacymatrix dataset is also MIT.
This server cannot be deployed
Maintenance
Related MCP Connectors
Read-only MCP server for The Quiet Protocol's engines, benchmarks, proof, and business data.
Agent-native MCP server over the public saagarpatel.dev corpus. Read-only, stateless.
Public, read-only MCP server for FarmNeural company facts, packages, and capabilities.
Public read-only MCP server for HODLXXI agent identity, trust, receipts, and verification.
Related MCP Servers
- AlicenseCqualityDmaintenanceAn extensible Model Context Protocol server that provides standardized access to social platform data (currently Farcaster) and onchain data for LLMs.1013MIT
- FlicenseAqualityDmaintenanceA Model Context Protocol server that provides read-only access to Datasette instances, enabling AI assistants to explore, query, and analyze data from Datasette databases through a standardized interface.52-
- FlicenseNot gradedqualityDmaintenanceA Model Context Protocol server that exposes SQL Server metadata and read-only query execution as a structured HTTP API, with safety validation and allowlist policy enforcement.1-
- AlicenseBqualityAmaintenanceA portable, read-only Model Context Protocol server for turning observability data into bounded evidence that AI agents can inspect safely.7Apache 2.0