Meilisearch

Meilisearch is a retrieval layer for RAG applications, combining vector storage, hybrid keyword and semantic search, and context formatting in one index. It ships with LangChain, LlamaIndex and MCP integrations and connects to major LLM providers.

Meilisearch positions itself as the retrieval foundation for AI applications, the layer that sits between your own data and whatever language model is generating answers. Rather than treating search as a separate concern, the platform bundles vector storage, hybrid retrieval, and context formatting into one service. Embeddings live alongside your documents inside Meilisearch, so a separate vector database is not part of the setup. The vendor says this stays fast at millions of documents, and that hybrid queries return results in under 50 milliseconds. The workflow follows the standard RAG pattern. A user asks a question in natural language, Meilisearch runs a hybrid search that blends keyword matching with semantic similarity to pull the most relevant documents, and the LLM then writes a response grounded in that retrieved content. A context formatting layer lets engineers choose exactly which fields and metadata reach the prompt, which keeps noise out of the model's input. Related features include a similar documents endpoint for "more like this" use cases and the ability to tune the keyword to semantic mix. This is a developer tool, not a no-code product. The target user is an engineering team building a RAG pipeline, an internal assistant, or an agent that needs live access to proprietary content. Official retriever integrations exist for LangChain in both Python and JavaScript, plus a vector store for LlamaIndex, and an MCP server lets Claude and other tool-using agents query the index. The site claims native support across 58 models from 13 providers, including OpenAI, Anthropic, Google, Mistral, Cohere, DeepSeek, Hugging Face, Ollama, Together AI, Fireworks, Cloudflare and AWS Bedrock. Compared with assembling a standalone vector database next to a separate keyword search engine, Meilisearch's pitch is consolidation: one index handles both retrieval modes, and the ecosystem hooks are already written. Teams that only need pure vector similarity, or that want a fully managed RAG service without operating an index, will find lighter options. The page lists Hugging Face, Louis Vuitton, Renault Group, Sky News, Bookshop.org and Agora among customers.