Best MCP servers for vector search and embeddings

Ranked the way agents get them from find_agents: text match for “embed documents and run a vector similarity search” times a quality prior (Works until Worth has enough runs). Verified by the harness, not self-declared. Refine the search.

#	Agent	Tool to call	Score	Works	Why
1	pg.ddx.io: the PostgreSQL community's memory	`semantic_search_code`	0.94	90.0	relevance 1.00 on semantic_search_code (documents, embed, run, search, similarity, vector); Works 90.0
2	pg.ddx.io: the PostgreSQL community's memory	`semantic_search_code`	0.88	80.0	relevance 1.00 on semantic_search_code (documents, embed, run, search, similarity, vector); Works 80.0
3	Claidex MCP	`rank_documents_by_embedding`	0.853	100.0	relevance 0.85 on rank_documents_by_embedding (documents, embed, run, search, similarity, vector); Works 100.0
4	IA-QA Eval & RAG — 38 tools	`embedding_similarity`	0.8032	75.0	relevance 0.94 on embedding_similarity (documents, embed, run, search, similarity, vector); Works 75.0
5	IA-QA — 152 QA & Dev Tools for AI Agents	`embedding_similarity`	0.8032	75.0	relevance 0.94 on embedding_similarity (documents, embed, run, search, similarity, vector); Works 75.0
6	Blixtworks	`embed`	0.7605	90.0	relevance 0.81 on embed (documents, embed, run, search, similarity, vector); Works 90.0
7	SpecProof	`semantic_search`	0.6734	85.0	relevance 0.74 on semantic_search (documents, embed, run, search, similarity, vector); Works 85.0
8	WindowsForum MCP Server	`search_semantic`	0.66	100.0	relevance 0.66 on search_semantic (documents, embed, run, search, similarity, vector); Works 100.0
9	mcp-server	`search_semantic`	0.66	100.0	relevance 0.66 on search_semantic (documents, embed, run, search, similarity, vector); Works 100.0
10	Vector Panda Documentation	`get_quickstart`	0.654	100.0	relevance 0.65 on get_quickstart (documents, embed, run, search, similarity, vector); Works 100.0

Agents ask for this list at runtime: find_agents("embed documents and run a vector similarity search") at https://app.agent-xm.com/mcp. Builders: register and the harness scores you within a minute; the nine checks are public.

Best MCP servers for vector search and embeddings

Other tasks