Official (Vendor)Active· 24d★ 4.3by Weaviate

Weaviate MCP

Inserts objects into Weaviate and runs hybrid search queries against vector collections.

View repo

Safety & Trust

Delv Safety Grade: B

Score 72/100 · assessed 2026-04-28

Maintainer85

Permissions65

Supply chain45

Transparency80

Incidents100

Weaviate's official MCP server connects Claude to their vector database for hybrid search and object insertion. The vendor is legitimate and well-established in the vector database space, with active maintenance and clear documentation. However, the supply chain story is weaker: you must clone and build from source using Make rather than installing from a package registry, and there's no versioned distribution. The permissions are moderately scoped—it writes to a remote database and reads API credentials from environment variables, which is standard for database tooling but does grant persistent write access. No security incidents are known. The hybrid search capability (vector plus keyword) is genuinely useful for RAG workflows, and being official means it tracks Weaviate's API changes. Transparency is good with open source code and adequate docs, though the install process adds friction and supply-chain verification burden.

Lethal Trifecta (prompt-injection exposure)

ONE OF THREE

Private dataYes

Reads secrets, credentials, private files

Untrusted inputNo

Ingests web pages, PRs, issues, emails

External commsNo

Can send data outbound

Same.

Green flags

Official vendor implementation from Weaviate team
Open source with visible commit history and issue tracker
Hybrid search combines vector similarity with keyword matching
Scoped to single database vendor, not arbitrary network access
No known security incidents or credential leaks

Red flags

No package registry distribution, must clone and build from source
Requires API key in environment, credential exposure risk if misconfigured
Write access to remote database with no apparent scoping or sandboxing
Make-based build process adds supply-chain verification burden

Permissions requested

Outbound networkAccess secretsDB readDB write

Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

make build

Env vars needed: WEAVIATE_URLWEAVIATE_API_KEY

Review

Weaviate MCP bridges Claude to Weaviate's vector database, letting you insert objects and run hybrid searches without leaving the chat window. It's the official server from Weaviate themselves, which means it tracks their API conventions and gets updated when the platform shifts. I'd reach for this when building RAG pipelines where Claude needs to query a knowledge base mid-conversation. The hybrid search is the standout feature: it combines vector similarity with keyword matching, so you get semantic understanding plus exact-match fallback. That's genuinely useful when users ask vague questions that need both conceptual and literal matches. The insert tool lets you add objects on the fly, which works well for chat memory or dynamically building a knowledge graph as a conversation unfolds. Setup requires a Weaviate instance (cloud or self-hosted), a URL, and an API key. The server itself is a make build away, but you'll need to wire those environment variables into your Claude Desktop config. Once running, you can insert JSON objects with properties and run searches that return ranked results with scores. The search tool accepts a collection name, query text, and optional filters, which covers most retrieval scenarios. Quirks: this is a thin wrapper around Weaviate's API, so you need to understand Weaviate's schema model. If your collection doesn't exist or your properties don't match the schema, you'll get cryptic errors. The server doesn't handle schema creation or migration, it assumes your collections are already configured. That's fine for production setups but adds friction if you're prototyping. Also, no batch operations beyond what the insert tool supports, so ingesting large datasets means scripting outside Claude. Skip this if you're not already using Weaviate or don't have vector search needs. It's not a general-purpose database tool, it's specifically for semantic retrieval. If you just need key-value storage or SQL queries, look elsewhere. But if you're running a Weaviate instance and want Claude to query it intelligently, this is the cleanest path.

Verdict

Install if you're already running Weaviate and need Claude to search or populate collections during conversations. The hybrid search is genuinely useful for RAG and semantic retrieval. Skip if you don't have vector search needs or aren't willing to manage Weaviate schemas separately.

Good at

Hybrid search combines vector similarity with keyword matching, covering both semantic and exact-match retrieval in one query.
Official vendor support means it tracks Weaviate API changes and gets maintained alongside the platform.
Insert tool lets you build chat memory or knowledge graphs dynamically during conversations.
Works with both Weaviate Cloud and self-hosted instances, so you control where your data lives.

Watch out

Requires an existing Weaviate instance with pre-configured schemas, no schema creation or migration tools included.
Error messages can be opaque if your collection or property names don't match the Weaviate schema.
No batch insert beyond single-object operations, so large ingestion tasks need external scripting.
Setup assumes familiarity with Weaviate's data model, not beginner-friendly if you're new to vector databases.

Use cases

vector search
RAG pipelines
chat memory
semantic indexing

Getting started

1. Run `make build` in the cloned repo to compile the server binary. 2. Add the server to your Claude Desktop config with `WEAVIATE_URL` and `WEAVIATE_API_KEY` environment variables pointing to your Weaviate instance. 3. Restart Claude Desktop and ask it to list available tools, you should see `insert-object` and `hybrid-search`. 4. Test by asking Claude to search a collection you've already created in Weaviate, providing the collection name and a query string. 5. Watch out: the server won't create collections or validate schemas, so make sure your Weaviate setup is ready before you start inserting objects.

Works with

Claude DesktopClaude CodeCursor