Hybrid Search & AI Search for Solr — Guides

Can AI agents like Claude or Cursor use Opensolr?

Yes. Opensolr ships an official MCP (Model Context Protocol) server, listed in the official MCP Registry under our domain-verified com.opensolr namespace. Any MCP-compatible client — Claude Desktop, Claude Code, Cursor, Windsurf — gets your search as native tools with one config entry:

uvx opensolr-mcp

The agent receives 8 tools: hybrid BM25 + kNN search, document indexing with server-side embeddings, grounded RAG answers from your own content, and index management. There is nothing to build — the chat is the UI: tell your agent "index our FAQ, then find everything about refunds" and it just happens.

Try it right now, without an account. There is a public demo account: email mcp@opensolr.com, API key 420b8b23e7b12dc8ab838932145a5065. The index mcp_demo_d1__dense is preloaded with 300 news articles, so search and grounded answers work the moment you connect, and you can create your own indexes on the account.

Anything you create there is deleted after 3 days. The account is shared with everyone — other people can change or delete your index, and you can do the same to theirs, so never put anything real in it. The limits are per index and deliberately small: 200 MB of bandwidth and 50 MB of disk. For an index that is private and stays put, start a free 15-day trial — no card — and swap in your own two values.

Learn more: MCP server page · MCP Registry listing · Source on GitHub

Read Full Answer

Does Opensolr integrate with Haystack?

Yes. Opensolr plugs into Haystack pipelines as a DocumentStore with a hybrid retriever:

pip install opensolr-haystack

A normal Haystack pipeline needs two embedder components — one for documents, one for queries. With OpensolrDocumentStore it needs none: everything embeds server-side on our GPU infrastructure, and OpensolrHybridRetriever fuses BM25 + kNN scores natively on managed Apache Solr 9. The full DocumentStore protocol is supported — duplicate policies, standard filters, Secret-based credentials, and serialization for saved pipelines.

Try it right now, without an account. There is a public demo account: email mcp@opensolr.com, API key 420b8b23e7b12dc8ab838932145a5065. The index mcp_demo_d1__dense is preloaded with 300 news articles, so search and grounded answers work the moment you connect, and you can create your own indexes on the account.

Anything you create there is deleted after 3 days. The account is shared with everyone — other people can change or delete your index, and you can do the same to theirs, so never put anything real in it. The limits are per index and deliberately small: 200 MB of bandwidth and 50 MB of disk. For an index that is private and stays put, start a free 15-day trial — no card — and swap in your own two values.

Learn more: Haystack integration page · PyPI package · Source on GitHub

Read Full Answer

Does Opensolr work with LangChain?

Yes. Opensolr is a native LangChain vector store, listed in the official LangChain integrations directory. Install it with:

pip install langchain-opensolr

You get OpensolrVectorStore and OpensolrEmbeddings: texts and queries are embedded server-side on our GPU infrastructure (multilingual, 1024 dimensions) — no OpenAI key, no local model — and search runs hybrid BM25 + kNN, fused per document with a tunable semantic↔lexical balance. It is the shortest vector-store setup in the LangChain directory: an API key and an index name. vs.as_retriever() drops it into any chain, agent, or RAG tutorial.

Try it right now, without an account. There is a public demo account: email mcp@opensolr.com, API key 420b8b23e7b12dc8ab838932145a5065. The index mcp_demo_d1__dense is preloaded with 300 news articles, so search and grounded answers work the moment you connect, and you can create your own indexes on the account.

Anything you create there is deleted after 3 days. The account is shared with everyone — other people can change or delete your index, and you can do the same to theirs, so never put anything real in it. The limits are per index and deliberately small: 200 MB of bandwidth and 50 MB of disk. For an index that is private and stays put, start a free 15-day trial — no card — and swap in your own two values.

Learn more: LangChain integration page · PyPI package · Source on GitHub

Read Full Answer