qdrant-llamaindex-mcp-server

Provides a Model Context Protocol (MCP) server exposing tools to read (and optionally write) documents stored in a Qdrant vector database that were indexed by LlamaIndex. It supports dynamic collection selection at runtime and attempts to adapt to various LlamaIndex payload/content field formats. For embeddings, it detects the embedding model per collection and embeds queries using a whitelisted set of allowed models, otherwise falling back to a default model.

Evaluated Apr 04, 2026 (27d ago)
Homepage ↗ Repo ↗ Ai Ml mcp qdrant llamaindex vector-database retrieval embeddings python fastmcp
⚙ Agent Friendliness
60
/ 100
Can an agent use this?
🔒 Security
48
/ 100
Is it safe for agents?
⚡ Reliability
28
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
78
Documentation
70
Error Messages
0
Auth Simplicity
75
Rate Limits
20

🔒 Security

TLS Enforcement
55
Auth Strength
45
Scope Granularity
15
Dep. Hygiene
60
Secret Handling
70

Reads are supported with QDRANT_READ_ONLY=true by design; write tools are gated behind QDRANT_READ_ONLY=false (good safety default). There is an embedding model whitelist with fallback behavior, reducing risk of accidental large model downloads. However, the README does not document MCP-level authentication/authorization, nor rate limiting, nor explicit guidance for securing SSE/HTTP transports. QDRANT_API_KEY is used (but secret logging behavior is not described). TLS enforcement is not specified for Qdrant connectivity (depends on QDRANT_URL).

⚡ Reliability

Uptime/SLA
0
Version Stability
40
Breaking Changes
35
Error Recovery
35
AF Security Reliability

Best When

You run an MCP-capable LLM app/agent that needs read-only or controlled write access to LlamaIndex data in Qdrant and you want automatic handling of differing payload schemas and embedding model selection.

Avoid When

You cannot restrict or secure access to the server/tooling (especially if write tools are enabled), or you require strict guarantees about embedding model provenance and resource usage while the embedding whitelist is permissive or disabled.

Use Cases

  • Semantic search and document retrieval over LlamaIndex-ingested content stored in Qdrant
  • Fetching points/documents and browsing collection contents via MCP tools
  • Building LLM applications/agents that need standardized tool access to Qdrant-backed knowledge bases
  • Debugging/inspecting Qdrant collections used by LlamaIndex (counts, sample points, collection details)

Not For

  • Exposing a public, general-purpose API for untrusted clients without additional network/auth controls (it is primarily an MCP server)
  • Use cases requiring fine-grained authorization per user/tenant
  • Operations needing full control over embedding models beyond the provided whitelist (unless whitelist is disabled)
  • High-scale production use without confirming operational robustness (rate limits, error handling details, and model loading behavior are not fully specified in the provided README)

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

Methods: Qdrant API key via QDRANT_API_KEY (passed through from environment) MCP transport security not specified in README; server is configured via FastMCP transport (stdio/sse/streamable-http)
OAuth: No Scopes: No

Authentication is primarily with Qdrant using QDRANT_API_KEY; MCP client authentication/authorization is not described. If exposing SSE/HTTP transports remotely, you should add network-level protection because the README does not specify MCP auth controls.

Pricing

Free tier: No
Requires CC: No

Open-source package; costs depend on Qdrant deployment and any embedding model downloads/runtime usage.

Agent Metadata

Pagination
offset/limit style for listing IDs/scrolling points (qdrant-list-document-ids, qdrant-scroll-points) plus limit default values
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • Write tools may be enabled when QDRANT_READ_ONLY=false; ensure you keep read-only mode on for safety with autonomous agents.
  • Embedding model whitelist can block collections; the server falls back to EMBEDDING_MODEL when a collection’s model is not allowed, which may produce retrieval quality changes.
  • Collection names are supplied dynamically by MCP clients; agents must pass correct collection_name for each call.
  • If using vector search by raw vector (qdrant-search-by-vector), clients must supply correctly shaped vectors (array of floats) matching the target collection’s vector config.

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for qdrant-llamaindex-mcp-server.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-04-04.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
Community Powered