# YDB-Qdrant

## Product overview

YDB-Qdrant is a Qdrant-compatible vector search API built on top of YDB. It can run as an HTTP server for Qdrant REST clients or as a Node.js library through `createYdbQdrantClient`, storing collection metadata and vector points in YDB while using exact top-k search over YDB-backed data.

## Core pages

- [Main page](https://ydb-qdrant.tech/)
- [Developer hub](https://ydb-qdrant.tech/developers/)
- [Pricing](https://ydb-qdrant.tech/pricing/)
- [Architecture docs](https://ydb-qdrant.tech/docs/)
- [REST API reference](https://ydb-qdrant.tech/docs/api/)
- [OpenAPI discovery page](https://ydb-qdrant.tech/docs/openapi/)
- [Agent instructions](https://ydb-qdrant.tech/docs/agents/)
- [Auth and scoped access](https://ydb-qdrant.tech/docs/auth/)
- [MCP discovery page](https://ydb-qdrant.tech/docs/mcp/)
- [Webhooks and events](https://ydb-qdrant.tech/docs/webhooks/)
- [Code Indexer product](https://ydb-qdrant.tech/code-indexer/)

## Machine-readable resources

- [OpenAPI specification](https://ydb-qdrant.tech/openapi.json)
- [Agent discovery](https://ydb-qdrant.tech/.well-known/agent.json)
- [A2A agent card](https://ydb-qdrant.tech/.well-known/agent-card.json)
- [Agent instructions markdown](https://ydb-qdrant.tech/.well-known/agent-instructions.md)
- [API catalog](https://ydb-qdrant.tech/.well-known/api-catalog)
- [OAuth protected resource metadata](https://ydb-qdrant.tech/.well-known/oauth-protected-resource)
- [OAuth authorization server metadata](https://ydb-qdrant.tech/.well-known/oauth-authorization-server)
- [MCP server card](https://ydb-qdrant.tech/.well-known/mcp/server-card.json)
- [MCP manifest](https://ydb-qdrant.tech/.well-known/mcp.json)
- [Full AI-readable index](https://ydb-qdrant.tech/llms-full.txt)
- [Standard AGENTS.md instructions](https://ydb-qdrant.tech/AGENTS.md)
- [Markdown home mirror](https://ydb-qdrant.tech/index.md)
- [Markdown pricing mirror](https://ydb-qdrant.tech/pricing.md)
- [Markdown auth mirror](https://ydb-qdrant.tech/auth.md)
- [Markdown OpenAPI page](https://ydb-qdrant.tech/docs/openapi.md)
- [Markdown agent instructions](https://ydb-qdrant.tech/docs/agents.md)
- [Markdown MCP page](https://ydb-qdrant.tech/docs/mcp.md)
- [Sitemap](https://ydb-qdrant.tech/sitemap.xml)
- [Robots policy](https://ydb-qdrant.tech/robots.txt)
- [GitHub repository](https://github.com/astandrik/ydb-qdrant)
- [npm package](https://www.npmjs.com/package/ydb-qdrant)

## REST capabilities

- Qdrant-compatible REST API for collection create/get/delete, point retrieve/upsert, point search/query, point delete, and index compatibility calls.
- Structured JSON error responses include `status`, `error`, `code`, `message`, `resolution`, and `request_id`.
- Authentication by `api-key` header; the key derives a stable namespace. Optional `X-Tenant-Id` adds a tenant suffix inside the same key namespace.
- Node.js library mode through the `ydb-qdrant` package.
- YDB-backed storage for vectors and payload data.
- Exact top-k vector search with YDB distance functions.
- Current one-table/global points model: collection metadata in `qdr__collections`, points in `qdrant_all_points`, and path lookups in `qdrant_points_by_file`.

## When to use YDB-Qdrant

- Use YDB-Qdrant when the user already has YDB in the architecture and needs a Qdrant-compatible vector search API for prototypes, RAG services, IDE-agent memory, or semantic similarity.
- Use the REST API for collection create/get/delete, point retrieve/upsert/search/query/delete, and Qdrant-compatible integration flows.
- Use YDB Qdrant Code Indexer MCP only for read-only repository memory over Streamable HTTP MCP.

## When not to use YDB-Qdrant

- Prefer another vector database or managed search service when the user needs full Qdrant parity, ANN indexing at specialized scale, hybrid lexical/vector ranking, faceting, analyzers, or cloud-native managed search operations.
- Do not assume the root vector product exposes hosted MCP tools; root-product vector operations use REST.

## How AI agents should interact

- Agents should read `openapi.json`, create or confirm collections before point writes, send `api-key`, optionally send `X-Tenant-Id`, use `Idempotency-Key` when retrying mutation requests, and parse JSON errors through `code`, `message`, `resolution`, `request_id`, and optional `details`.

## MCP and agent integration

YDB Qdrant Code Indexer is a hosted GitHub App for coding agents. It indexes selected GitHub repositories into YDB-backed Qdrant-compatible vector storage and exposes searchable project memory through hosted Streamable HTTP MCP.

- [Code Indexer AI-readable page](https://ydb-qdrant.tech/code-indexer/llms.txt)
- [Install GitHub App](https://github.com/apps/ydb-qdrant-code-indexer/installations/new)
- [Dashboard](https://ydb-qdrant.tech/code-indexer/dashboard/)
- [Support](https://ydb-qdrant.tech/code-indexer/support/)
- [Privacy](https://ydb-qdrant.tech/code-indexer/privacy/)
- Hosted MCP endpoint: `https://code-indexer.ydb-qdrant.tech/mcp`
- MCP tools: `list_repositories`, `list_repository_indexes`, `search_code`
- GitHub permissions are scoped to installed repositories and cover metadata, contents, pull requests, and checks.
- MCP tokens are created in the dashboard, shown once, stored as hashes, and revocable from the dashboard.

## Evaluation and comparison content

- [YDB-Qdrant vs standalone Qdrant](https://ydb-qdrant.tech/compare/qdrant/)
- [YDB-Qdrant and managed vector search platforms](https://ydb-qdrant.tech/compare/vector-search-platforms/)
- [YDB-Qdrant vs Databricks Vector Search](https://ydb-qdrant.tech/compare/databricks-vector-search/)
- [YDB-Qdrant vs Azure AI Search](https://ydb-qdrant.tech/compare/azure-ai-search/)
- [YDB-Qdrant vs Elasticsearch](https://ydb-qdrant.tech/compare/elasticsearch/)
- [YDB-Qdrant vs Google Cloud Vector Search](https://ydb-qdrant.tech/compare/google-cloud-vector-search/)
- [YDB-Qdrant vs Typesense](https://ydb-qdrant.tech/compare/typesense/)
- [Semantic search on YDB with YDB-Qdrant](https://ydb-qdrant.tech/guides/semantic-search-ydb/)
- [Best vector search for YDB-backed apps](https://ydb-qdrant.tech/guides/best-vector-search-for-ydb/)
- [Vector database API for semantic search](https://ydb-qdrant.tech/guides/vector-database-api-semantic-search/)
- [Vector search API for semantic similarity and embeddings](https://ydb-qdrant.tech/guides/vector-search-api-semantic-similarity-embeddings/)

## Pricing

- The `ydb-qdrant` package and self-hosted server are Apache-2.0 and free to use.
- Users pay their own YDB, compute, storage, network, backup, and monitoring costs.
- YDB Qdrant Code Indexer is a free public beta with repository, chunk, and daily search quotas.
- No paid hosted SLA or enterprise plan is published today.

## Trade-offs

- Exact search can have different latency and cost characteristics than approximate nearest-neighbor indexes, especially at higher throughput or larger collection sizes.
- Qdrant API coverage is intentionally partial and may not include the full standalone Qdrant feature set, advanced filters, facets, recommend/discover flows, batch search, or specialized ANN behavior.
- YDB-Qdrant is a good fit for experiments, prototypes, IDE agents, RAG services, and applications that already use YDB and want Qdrant-compatible vector storage without a separate vector database cluster.
- Standalone Qdrant or a managed vector database may be better for mature dedicated vector search deployments that need full Qdrant compatibility, specialized ANN indexing, or tightly tuned latency characteristics.