Question 1

Do I need Obsidian installed?

Accepted Answer

No. enquire-mcp reads .md, .canvas, and .pdf files directly from disk and works against any Obsidian-format vault — the Obsidian desktop app does not need to be running.

Question 2

Will it write to my vault?

Accepted Answer

Not unless you pass --enable-write. All 7 write tools are gated behind that flag, and the destructive ones support a dry_run preview.

Question 3

Is my data sent anywhere?

Accepted Answer

Only on `enquire-mcp install-model`, which downloads ONNX model weights from HuggingFace one time. Serve mode makes zero outbound HTTP calls — embeddings and the reranker run on CPU locally, so your vault content never leaves your machine.

Question 4

What is the query performance?

Accepted Answer

Cold-build FTS5 is ~5s per 1k notes (~30s per 50k). A BM25 query is always under 100ms, and HNSW top-10 vector search is sub-10ms at any scale. Serve cold-start is ~50ms with HNSW persistence.

Question 5

What languages are supported?

Accepted Answer

The default paraphrase-multilingual-MiniLM-L12-v2 model covers 50+ languages with a multilingual cross-encoder, validated end-to-end on bilingual Russian + English vaults. CJK/Thai/Khmer tokenization uses Intl.Segmenter.

Question 6

Can I run it remotely?

Accepted Answer

Yes — `serve-http` exposes the same server over Streamable HTTP. Front it with Tailscale Funnel or Cloudflare Tunnel for HTTPS. It works with claude.ai web, ChatGPT custom GPTs, Cursor HTTP mode, and mobile MCP clients.

enquire-mcp API reference - v3.11.6-rc.2

Modules

Settings

On This Page