# GigaRAG > Deploy production-ready retrieval pipelines, hybrid search, embeddings, and contextual memory through one API. Millisecond retrieval, no vector database to manage. GigaRAG is retrieval infrastructure for AI applications. It provides managed indexing, embeddings, hybrid search (semantic + BM25 fused with reciprocal rank fusion), reranking, knowledge-graph memory, and agentic retrieval through a single REST API and first-party SDKs. There is no vector database to provision, size, or patch. ## Key facts - Product category: retrieval-augmented generation (RAG) infrastructure, vector search, AI memory. - Interfaces: REST API, SDKs for Node.js, Python, Java, Kotlin, and Swift, and an MCP integration. - Embeddings: bring your own from any provider, or use ours. Vectors up to 2048 dimensions. Models can differ per index. - Retrieval: hybrid semantic + keyword search with RRF fusion and reranking, typically under 100ms at p95. - Scaling: storage and retrieval scale automatically; billing follows usage rather than provisioned capacity. - Security: AES-256 at rest, TLS in transit, per-tenant isolation. Customer content is never used to train models. - Pricing: free tier available, no credit card required to start. Paid plans renew automatically and are non-refundable, because capacity is allocated for the full period. ## Pages - [Home](https://gigarag.com/): product overview, features, integrations, and developer quickstart. - [API reference](https://gigarag.com/docs/api.md): every REST endpoint with its fields and examples, authentication, errors, and the MCP endpoint, as plain markdown. - [Connect an agent](https://gigarag.com/connect): how to give an AI tool access to a GigaRAG workspace over MCP, with a setup guide per client at https://gigarag.com/connect/. 99 clients are covered, including Claude Code, Cursor, ChatGPT, Claude, VS Code. - [Blog](https://gigarag.com/blog): guides on retrieval-augmented generation, hybrid search, and embeddings. An RSS feed is at https://gigarag.com/blog/feed.xml. - [Terms of Service](https://gigarag.com/terms): account, billing, refund, and liability terms. - [Privacy Policy](https://gigarag.com/privacy): what is collected, how it is protected, and what is never done with it. ## Frequently asked questions ### What is RAG infrastructure? Retrieval-augmented generation infrastructure is the layer that stores your content, embeds it, and fetches the most relevant passages at query time so a model answers from your data instead of guessing. ### Which embedding models are supported? Bring your own from OpenAI, Gemini, Claude, or any open-source model, or use ours. Vectors up to 2048 dimensions are supported, and you can mix models across indexes. ### Do I need to manage vector databases manually? No. Indexes are provisioned, sharded, and scaled for you. There are no clusters to size and no nodes to patch. ### What data sources can I connect? Files, object storage, databases, and SaaS apps via our connectors, plus anything you push through the ingestion API or an MCP server. ### How fast is retrieval performance? Typical hybrid queries return in well under 100ms at p95, including reranking, from the region closest to your workload. ### Does the platform support hybrid search? Yes. Semantic search and BM25 run together and are fused with reciprocal rank fusion, so you get both exact keyword matches and conceptual recall. ### Can I integrate this with my existing AI stack? Yes. Use the REST API, an official SDK, or the MCP integration to drop retrieval into the agent framework and tooling you already run. ### Is the infrastructure scalable? Storage and retrieval scale automatically with your workload. You pay for what you use rather than for provisioned capacity. ### How secure is stored data? Data is encrypted with AES-256 at rest and TLS in transit, isolated per tenant, and never used to train models. ### Do you provide APIs and SDKs? Yes — a documented REST API plus first-party SDKs for Node.js, Python, Java, Kotlin, and Swift. ## Contact Email hello@gigarag.com.