Vector search, memory and RAG APIs and MCP servers for AI agents

Vector databases, retrieval (RAG) services and long-term memory layers for agents.

9 vector search, memory and rag APIs an AI agent can call: auth method and header, official MCP server (9), free tier (8), docs links. Machine-readable JSON included. Agents: fetch catalog/databases-vector-memory.json instead of parsing this page. Last checked 2026-09-24.

Ratings and usage (★) fetched: 2026-09-24 16:54 UTC – 2026-09-24 17:02 UTC.

Service	What an agent can do	Auth	MCP server	Free tier	Details (with sources)	Notes	Proof (ratings, usage)
Agentset RAG API	Lets an agent ingest documents into managed RAG namespaces and run semantic search or cited question answering over them.	API key `Authorization: Bearer <API_KEY>`	official	Free plan (forever): 1,000 pages ingested and 10,000 retrievals per month, 60 requests/min per namespace	Base URL: https://api.agentset.ai Free plan: 1,000 pages; 10,000 retrievals per month ↗ Example call: search a namespace: POST https://api.agentset.ai/v1/namespace/{namespaceId}/search ↗ Unit price: Retrievals: 10,000/month on Free, unlimited on Pro ($49/mo, 10,000 pages included); ingestion $0.01 per additional page ↗	MCP server is local (stdio, npx @agentset/mcp) and needs AGENTSET_API_KEY plus a namespace ID. Open-source, self-hostable.	—
Chroma Cloud	Create databases and store/query vector embeddings for retrieval via REST/SDK or MCP server.	API key `x-chroma-token or Authorization header`	official	Trial: Starter plan $0/month base with $5 one-time free credits on signup, then usage-based pricing	Base URL: https://api.trychroma.com Free plan: Starter plan $0/month base with $5 in free credits, then usage-based, one-time ↗ Example call: query a collection by embedding: POST https://api.trychroma.com/api/v2/tenants/{tenant}/databases/{database}/collections/{collection_id}/query ↗ Unit price: Reads: $0.0075 per TiB queried + $0.09 per GiB returned; writes $2.50 per logical GiB; storage $0.33 per GiB-month ↗ MCP tools: 12: chroma_list_collections, chroma_create_collection, chroma_peek_collection, chroma_get_collection_info, chroma_get_collection_count, chroma_modify_collection … ↗		GitHub 599★ chroma-core/chroma-mcp (2026-09-24 16:54 UTC)
Letta	Create and run stateful agents with persistent memory blocks via REST API or hosted MCP server.	API key `Authorization: Bearer <api_key>`	official	Free: $0/month plan with up to 3 stateful agents and limited Letta Auto usage (bring your own API keys); API plan $20/month + usage	Base URL: https://api.letta.com Free plan: Free plan $0/month: up to 3 stateful agents, limited Letta Auto usage, bring your own API keys ↗ Example call: send a message to a stateful agent: POST https://api.letta.com/v1/agents/{agent_id}/messages ↗ Unit price: API plan $20/month + $0.10 per active agent/month + $0.00015 per second of tool execution; LLM tokens billed at model rates ↗ MCP tools: 6: list_agents, list_models, create_agent, send_agent_message, get_run, get_reply ↗		—
Mem0	Add, search, update, and delete long-term agent memories via REST API or hosted MCP server.	API key `Authorization: Token <api_key>`	official	free (Hobby) tier: 10K add requests, 1K retrievals/mo, 1 project	Base URL: https://api.mem0.ai Free plan: Hobby: 10,000 add requests and 1,000 retrieval requests per month ↗ Async jobs: POST /v3/memories/add/ returns event_id with status PENDING; check event status (MCP tool get_event_status) or use webhooks ↗ Example call: add memories from a conversation: POST https://api.mem0.ai/v3/memories/add/ ↗ MCP tools: 11: add_memory, search_memories, get_memories, get_memory, update_memory, delete_memory … ↗	Graph memory feature is gated to the Pro plan.	—
Pinecone	Upsert, query, and rerank vector embeddings in managed indexes via REST/SDK or MCP server.	API key `Api-Key header`	official	free tier (Starter plan, limited pods/reads)	Base URL: https://api.pinecone.io Free plan: Starter: up to 2 GB storage, 1M read units, 2M write units, 5 indexes; Assistant 1 GB storage, per month ↗ Rate limits: Starter: 1M read units and 2M write units per month per org; 100 query and 100 upsert requests/s per namespace; 429 when exceeded ↗ Example call: query an index for nearest vectors: POST https://{INDEX_HOST}/query ↗ Unit price: Standard plan: read units $16-$18 per million, write units $4-$4.50 per million, storage $0.33/GB-month; $50/month minimum ↗ MCP tools: 9: search-docs, list-indexes, describe-index, describe-index-stats, create-index-for-model, upsert-records … ↗	Two official MCP servers: Developer/Assistant remote MCP and a local variant.	GitHub 72★ pinecone-io/pinecone-mcp (2026-09-24 16:54 UTC)
Qdrant Cloud	Create clusters and store/search vector embeddings via REST/gRPC API, SDKs, or official MCP server.	API key `api-key header`	official	Free tier: single-node cluster with 0.5 vCPU, 1 GB RAM, 4 GB disk, no card; suspends after 1 week idle	Base URL: https://<cluster>.<region>.<provider>.cloud.qdrant.io:6333 Free plan: Free forever single-node cluster: 0.5 vCPU, 1 GB RAM, 4 GB disk ↗ Example call: query nearest points by vector: POST https://{cluster_host}:6333/collections/{collection_name}/points/query ↗ MCP tools: 2: qdrant-store, qdrant-find ↗	Free clusters delete after 4 weeks of inactivity if not reactivated.	GitHub 1,537★ qdrant/mcp-server-qdrant (2026-09-24 16:54 UTC)
Weaviate Cloud	Store and semantically search vector objects in a managed cluster via REST/GraphQL or MCP server.	API key `Authorization: Bearer <api_key>`	official	Free: always-free plan, 1 cluster, 100,000 objects, 1 GB memory, 10 GB disk, 1 collection, no card required	Base URL: https://<cluster-host>/v1 Free plan: Always free: 1 cluster, 100,000 objects, 1 GB memory, 10 GB disk, 1 collection; embeddings 2,000 req/day, no card ↗ Example call: semantic (nearText) search in a collection: POST https://{cluster-host}/v1/graphql ↗ MCP tools: 4: weaviate-collections-get-config, weaviate-tenants-list, weaviate-query-hybrid, weaviate-objects-upsert ↗	MCP endpoint built into the database at /v1/mcp: always on in Weaviate Cloud, disabled by default when self-hosted (writes need MCP_SERVER_WRITE_ACCESS_ENABLED).	GitHub 163★ weaviate/mcp-server-weaviate (2026-09-24 16:54 UTC)
Zep	Store and retrieve temporal knowledge-graph agent memory via REST API or a single MCP endpoint.	API key `Authorization: Api-Key <api_key> (REST API); MCP server uses OAuth 2.1 via the org's IdP`	official	Free tier: 10,000 credits per month, no rollover; variable rate limits	Base URL: https://api.getzep.com/api/v2 Free plan: 10,000 credits per month, no rollover; variable rate limits ↗ Rate limits: Free plan: variable rate limits depending on service-wide load ↗ Example call: add chat messages to a user's thread memory: POST https://api.getzep.com/api/v2/threads/{threadId}/messages ↗ Unit price: 1 credit per episode up to 350 bytes (+1 per extra 350 bytes); extra credits $25 per 10,000 on Flex ↗ MCP tools: 9: search_graph, get_user_summary, get_subgraph, get_node_neighbors, list_episodes, list_graphs … ↗	One MCP URL serves both Zep managed cloud and BYOC deployments.	—
Zilliz Cloud (Milvus)	Manage Milvus clusters and perform vector search/ingestion via REST API, SDKs, or MCP server.	API key `Authorization: Bearer <token>`	official	Free tier: serverless cluster, 5GB storage, 2.5M vCUs per month, up to 5 collections, no card; plus up to $200 one-time trial credits ($100 on signup + $100 on adding a card)	Base URL: https://api.cloud.zilliz.com/v2 Free plan: 5 GB storage, 2.5M vCUs per month, up to 5 collections; separately $200 free credits for evaluation, no card ↗ Example call: vector similarity search in a collection: POST {CLUSTER_ENDPOINT}/v2/vectordb/entities/search ↗ MCP tools: 16: list_projects, list_clusters, create_free_cluster, describe_cluster, suspend_cluster, resume_cluster … ↗	MCP server can even provision a new free-tier cluster on request.	GitHub 34★ zilliztech/zilliz-mcp-server (2026-09-24 16:54 UTC)

Vector search, memory and RAG APIs and MCP servers for AI agents

Other categories