How to Connect Qdrant to an AI Agent
Auth setup
1. Qdrant Cloud: create a free cluster at cloud.qdrant.io, generate an API key from the cluster dashboard. 2. Self-host: docker run -p 6333:6333 qdrant/qdrant (no auth by default — add API key via QDRANT__SERVICE__API_KEY env var). 3. MCP: npx @qdrant/mcp-server --url https://xyz.qdrant.io --api-key ***
Key facts
| Base URL | http://localhost:6333/ |
|---|
| API version | v1 (REST) + gRPC |
|---|
| Auth | API key authentication via Api-Key header (Qdrant Cloud) or open access (self-hosted by default, secure with a reverse proxy). Qdrant Cloud clusters expose HTTPS endpoints per cluster. Official MCP server: npx @qdrant/mcp-server. |
|---|
| Scopes | collection:read,collection:write (or cluster-wide with a cluster API key) |
|---|
| Request body | application/json |
|---|
| Pagination | Scroll API: POST /points/scroll with limit and offset (point_id). Search limit max 10,000. |
|---|
| Rate limit | No documented hard rate limit on self-hosted. Qdrant Cloud: depends on cluster tier — free tier ~100 req/s, paid tiers scale with cluster size. gRPC is preferred for bulk upserts (~5x faster than REST). |
|---|
| Error format | JSON: {"status":{"error":"Wrong input: Vector dimension error: expected dim 1536, got 768"},"time":0.001} |
|---|
Key endpoints
| Method | Path | Description |
|---|
PUT | /collections/{collection_name} | Create a collection with vector size, distance metric (Cosine/Dot/Euclid), and optional quantization |
GET | /collections/{collection_name} | Get collection info (vector count, status, config) |
PUT | /collections/{collection_name}/points | Upsert points (id, vector, payload). Supports batch via the 'batch' variant. |
POST | /collections/{collection_name}/points/search | Similarity search with optional payload filter and score threshold |
POST | /collections/{collection_name}/points/query | Unified query (vector + filter + prefetch for hybrid search) |
POST | /collections/{collection_name}/points/delete | Delete points by id or filter |
Quickstart
POST /collections/docs/points/search
Api-Key: {key}
Content-Type: application/json
{"vector":[0.1,0.2,...],"limit":5,"with_payload":true,"score_threshold":0.7}
Response: {"result":[{"id":42,"score":0.87,"payload":{"source":"docs/intro.md"}}]}
Agent pitfalls & tips
- Vector size MUST match the embedding model (1536 for text-embedding-3-small, 3072 for text-embedding-3-large, 768 for nomic-embed-text). Mismatched sizes produce errors on upsert.
- Use Cosine distance for normalized embeddings (most OpenAI/Anthropic models) and Dot for raw inner-product semantics.
- Payload filters combine with vector search — use them for tenant isolation, time filtering, or source restriction without re-indexing.
- Quantization (scalar or binary) reduces memory 4-32x with minimal recall loss — enable it for large collections.
- For hybrid search (dense + sparse), use the /points/query endpoint with prefetch — Qdrant supports BM25-style sparse vectors natively.
- Always set a score_threshold (typically 0.6-0.8 for cosine) to filter out irrelevant results — improves RAG quality significantly.
- Batch upserts via gRPC or REST's batch variant — single upserts are ~5-10x slower at scale.
- The official MCP server (npx @qdrant/mcp-server) exposes search, upsert, and collection management — preferred over raw REST for agents.
Source: curated by KanseiLink from official documentation (docs) and registry checks. Last reviewed: 2026-04-10. Specs change — verify against the official docs before production use.
Frequently Asked Questions
What is Qdrant's AEO score?
▼
Qdrant has an AEO score of 1.00 and is rated AAA (Best-in-class agent integration). AEO (Agent Engine Optimization) measures how well a SaaS service works with AI agents. Scores range from 0.00 to 1.00, with grades from AAA (best) to D (not agent-ready).
Is Qdrant AI-agent-ready?
▼
Qdrant is currently ✓ verified for AI agent use. It offers an official MCP (Model Context Protocol) server, which means AI agents can connect directly. For detailed connection guides, auth setup, and known pitfalls, use the KanseiLink MCP tool.
How does Qdrant compare to other Database services?
▼
In the Database category, Qdrant is rated AAA. KanseiLink evaluates services based on MCP availability, API quality, documentation, auth-guide clarity, and integration recipe availability (methodology published). Visit the full rankings at kansei-link.com to see how Qdrant compares.
How can I integrate Qdrant with an AI agent?
▼
The fastest way to integrate Qdrant with an AI agent is through KanseiLink MCP. Install it with: npx @kansei-link/mcp-server — then use the search_services and get_service_detail tools to get the current auth setup, endpoints, rate limits, and agent-specific tips. This data is kept fresh from registry checks, curated official-doc guides, and agent reports.
How do I authenticate with Qdrant?
▼
API key authentication via Api-Key header (Qdrant Cloud) or open access (self-hosted by default, secure with a reverse proxy). Qdrant Cloud clusters expose HTTPS endpoints per cluster. Official MCP server: npx @qdrant/mcp-server. Setup: 1. Qdrant Cloud: create a free cluster at cloud.qdrant.io, generate an API key from the cluster dashboard. 2. Self-host: docker run -p 6333:6333 qdrant/qdrant (no auth by default — add API key via QDRANT__SERVICE__API_KEY env var). 3. MCP: npx @qdrant/mcp-server --url https://xyz.qdrant.io --api-key ***
What are Qdrant's API rate limits?
▼
No documented hard rate limit on self-hosted. Qdrant Cloud: depends on cluster tier — free tier ~100 req/s, paid tiers scale with cluster size. gRPC is preferred for bulk upserts (~5x faster than REST).