Skip to main content
mcpSource-backed
Milvus logo

Milvus MCP Server

MCP server from Zilliz for connecting Claude to Milvus vector database collections, text search, vector search, hybrid search, inserts, deletes, indexes, collection loading, database switching, and collection metadata.

by Zilliz · submitted by oktofeesh1·added 2026-06-06·
Review first review before installing

Open the source and read safety notes before installing.

Citation facts

Source-backed facts for citing this resource, derived directly from the registry — also available as plain text for AI assistants.

Source URLs
https://github.com/zilliztech/mcp-server-milvus#readme, https://github.com/zilliztech/mcp-server-milvus
Brand
Milvus
Brand domain
milvus.io
Brand asset source
brandfetch
Safety notes
Milvus MCP can read collection metadata, query collections, and run text, vector, text-similarity, multi-vector, and hybrid searches., Write-capable tools can create collections, insert data, upsert data, delete entities, create indexes, bulk insert records, load collections, release collections, and switch databases., The README notes that the environment file has higher priority than command-line arguments, so stale or unexpected environment settings can silently change the target Milvus instance., SSE and Streamable HTTP modes can expose database operations over HTTP and should be network-restricted., Remote Milvus or Zilliz Cloud credentials should be scoped to the collections and databases Claude is allowed to access.
Privacy notes
Milvus collections can contain embeddings, sparse vectors, scalar fields, IDs, document chunks, metadata, image or multimodal references, query logs, and retrieval results that reveal sensitive project or user data., Milvus URI, tokens, database names, collection names, vector payloads, filter expressions, and retrieved records should stay out of prompts, issues, logs, screenshots, and committed files., Search results can include private source content that may be re-exposed in model transcripts or downstream tickets., HTTP transports, debug tools, query traces, failed-search artifacts, backups, and benchmark datasets need retention and access-control review.
Author
Zilliz
Submitted by
oktofeesh1
Claim status
unclaimed
Last verified
2026-06-06

Decision playbook

Review trust signals before you adopt

Signals are present but mixed. Use the checklist below to confirm the source and operational safety for your environment.

Compare context
Selected

0

Current score

63

Baseline

Delta

No baseline selected

No major trust-signal divergence detected in the current selection.

Source and provenance checks

Needs review

Confirm ownership and provenance before trusting install instructions.

  • Source link availableRequired

    Open the canonical repository and verify ownership.

    Done
  • Source provenance statusRequired

    Marked as source-backed.

    Done
  • Metadata reviewed

    No reviewed flag detected in metadata.

    Pending

Safety and privacy checks

Complete

Validate risk disclosures before installation or API wiring.

  • Safety notes presentRequired

    Review the listed safety guidance before running commands.

    Done
  • Privacy notes presentRequired

    Review data handling notes before connecting accounts or secrets.

    Done
  • Trust level risk gateRequired

    Trust level does not block evaluation.

    Done

Package and install checks

Needs review

Check package metadata and artifact integrity signals.

  • Install payload available

    Install or copy payload is available for review.

    Done
  • Package verification flag

    No package verification flag provided.

    Pending
  • Checksum metadata

    No checksum provided for downloaded artifact.

    Pending

Compare-driven decision checks

Needs review

Use compare context to validate trade-offs before adoption.

  • Compare tray has multiple entries

    Add at least one more entry to compare trust differences.

    Pending
  • Baseline comparison available

    No baseline peer selected yet.

    Pending
  • Diverging trust signals identified

    No major trust-signal divergence found.

    Pending

Setup at a glance

CLI install

Copy-ready — paste the snippet to get started.

20 minutes

Adoption plan

Balanced adoption plan

Current risk score 24/100. Use staged verification before broader rollout.

Risk 24

Pre-adoption checks

Validate source and review signals before any execution.

  • Confirm source provenanceRequired

    Source URL/provenance metadata is present.

    Done
  • Confirm metadata review state

    No review metadata found; increase manual validation.

    Pending
  • Verify install payload

    Install/config payload exists and can be inspected.

    Done

Security checks

Confirm safety, privacy, and package integrity signals.

  • Review safety notesRequired

    Safety notes are present.

    Done
  • Review privacy notesRequired

    Privacy notes are present.

    Done
  • Verify package integrity metadata

    No package verification/checksum metadata.

    Pending

Rollout

Adopt in controlled steps based on the selected plan.

  • Run in isolated sandbox firstRequired

    Use a constrained sandbox and observe behavior across multiple tasks.

    Pending
  • Roll out graduallyRequired

    Roll out to a small cohort before wider usage.

    Pending
  • Set monitoring and fallback

    Define rollback path and monitor errors after adoption.

    Pending

Evidence readiness

Evidence readiness matrix · balanced

Missing required evidence: Metadata review. Risk score 31.

Risk 31

Source provenance

Present

Source repository/provenance is listed.

Required in this preset

Metadata review

Missing

Review metadata is missing.

Required in this preset

Safety notes

Present

Safety notes are present.

Required in this preset

Privacy notes

Present

Privacy notes are present.

Optional in this preset

Package integrity

Missing

Package integrity metadata is missing.

Optional in this preset

Install payload

Present

Install payload is available.

Required in this preset

Required gaps: Metadata review

Decision timeline

Decision timeline · balanced

Blocking gaps: Check metadata review status. Risk 28.

Risk 28

triage

Confirm source provenanceRequired

Source/provenance metadata is available.

Done

triage

Check metadata review statusRequired

Review metadata is missing.

Pending

verify

Review safety notesRequired

Safety notes are available.

Done

verify

Review privacy notes

Privacy notes are available.

Done

verify

Validate package integrity metadata

Package integrity metadata is missing.

Pending

rollout

Verify install payload and commandsRequired

Install payload is available.

Done

Blockers: Check metadata review status

Prerequisite readiness

Prerequisite readiness

6 prerequisites to line up before setup. Have accounts and credentials ready first. Includes a review or approval gate.

0/6 ready
Account & credentials1Install & runtime2Review & approval2General120 minutes

Safety & privacy surface

Safety & privacy surface

5 safety and 4 privacy notes across 6 risk areas. Review closely: credentials & tokens, network access.

6 areas
  • SafetyExecution & processesMilvus MCP can read collection metadata, query collections, and run text, vector, text-similarity, multi-vector, and hybrid searches.
  • SafetyGeneralWrite-capable tools can create collections, insert data, upsert data, delete entities, create indexes, bulk insert records, load collections, release collections, and switch databases.
  • SafetyLocal filesThe README notes that the environment file has higher priority than command-line arguments, so stale or unexpected environment settings can silently change the target Milvus instance.
  • SafetyNetwork accessSSE and Streamable HTTP modes can expose database operations over HTTP and should be network-restricted.
  • SafetyCredentials & tokensRemote Milvus or Zilliz Cloud credentials should be scoped to the collections and databases Claude is allowed to access.
  • PrivacyData retentionMilvus collections can contain embeddings, sparse vectors, scalar fields, IDs, document chunks, metadata, image or multimodal references, query logs, and retrieval results that reveal sensitive project or user data.
  • PrivacyCredentials & tokensMilvus URI, tokens, database names, collection names, vector payloads, filter expressions, and retrieved records should stay out of prompts, issues, logs, screenshots, and committed files.
  • PrivacyExecution & processesSearch results can include private source content that may be re-exposed in model transcripts or downstream tickets.
  • PrivacyNetwork accessHTTP transports, debug tools, query traces, failed-search artifacts, backups, and benchmark datasets need retention and access-control review.

Safety notes

  • Milvus MCP can read collection metadata, query collections, and run text, vector, text-similarity, multi-vector, and hybrid searches.
  • Write-capable tools can create collections, insert data, upsert data, delete entities, create indexes, bulk insert records, load collections, release collections, and switch databases.
  • The README notes that the environment file has higher priority than command-line arguments, so stale or unexpected environment settings can silently change the target Milvus instance.
  • SSE and Streamable HTTP modes can expose database operations over HTTP and should be network-restricted.
  • Remote Milvus or Zilliz Cloud credentials should be scoped to the collections and databases Claude is allowed to access.

Privacy notes

  • Milvus collections can contain embeddings, sparse vectors, scalar fields, IDs, document chunks, metadata, image or multimodal references, query logs, and retrieval results that reveal sensitive project or user data.
  • Milvus URI, tokens, database names, collection names, vector payloads, filter expressions, and retrieved records should stay out of prompts, issues, logs, screenshots, and committed files.
  • Search results can include private source content that may be re-exposed in model transcripts or downstream tickets.
  • HTTP transports, debug tools, query traces, failed-search artifacts, backups, and benchmark datasets need retention and access-control review.

Prerequisites

  • Python 3.10 or newer.
  • uv installed for the README's recommended run path.
  • Running local or remote Milvus instance.
  • Milvus URI, token, and database name selected for the target environment.
  • Review of whether stdio, SSE, or Streamable HTTP mode is appropriate for the MCP client.
  • Human approval policy for collection creation, inserts, upserts, deletes, index changes, database switching, and collection loading or release operations.

Schema details

Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
20 minutes
Difficulty
advanced
Full copyable content
{
  "mcpServers": {
    "milvus": {
      "command": "uv",
      "args": [
        "--directory",
        "mcp-server-milvus",
        "run",
        "src/mcp_server_milvus/server.py",
        "--milvus-uri",
        "<milvus-uri>"
      ]
    }
  }
}

About this resource

Content

Milvus MCP Server is a Zilliz-maintained Model Context Protocol server for Milvus vector database workflows. It gives Claude and other MCP clients access to Milvus collections, text search, vector search, hybrid search, query filters, collection metadata, index operations, inserts, deletes, database switching, and collection load or release operations.

The README documents stdio, SSE, and Streamable HTTP modes. For local Claude or Cursor use, stdio is the safest starting point because the server process stays attached to the MCP client instead of exposing HTTP endpoints.

Source Review

These sources were reviewed on 2026-06-06. The public Milvus website and docs blocked command-line URL checks, and PyPI metadata for the package name points to a different homepage, so this entry intentionally uses only reachable GitHub source evidence from zilliztech/mcp-server-milvus.

Features

  • Zilliz-maintained MCP server for Milvus.
  • Stdio, SSE, and Streamable HTTP modes.
  • Collection listing and collection information tools.
  • Text search, vector search, text-similarity search, multi-vector search, and hybrid search.
  • Filtered query operations.
  • Collection creation with schema and vector configuration.
  • Insert, upsert, bulk insert, and delete-entity operations.
  • Index creation and index information tools.
  • Collection stats, loading progress, load, and release tools.
  • Database listing and database switching.
  • Environment-file and command-line configuration for Milvus URI, token, and database.

Installation

Clone the repository and run the server with uv as documented in the README:

git clone https://github.com/zilliztech/mcp-server-milvus.git
cd mcp-server-milvus
uv run src/mcp_server_milvus/server.py --milvus-uri <milvus-uri>

For an MCP client configuration that launches from a checked-out copy:

{
  "mcpServers": {
    "milvus": {
      "command": "uv",
      "args": [
        "--directory",
        "mcp-server-milvus",
        "run",
        "src/mcp_server_milvus/server.py",
        "--milvus-uri",
        "<milvus-uri>"
      ]
    }
  }
}

Review the target URI, token, database, and environment file before starting the server. The README says the environment file takes priority over command-line arguments.

Use Cases

  • Ask Claude to list Milvus collections before selecting a retrieval source.
  • Inspect collection metadata and stats while debugging ingestion.
  • Run vector, text, or hybrid searches with explicit limits and output fields.
  • Query collections with filter expressions.
  • Create an experimental collection for a local evaluation.
  • Insert or upsert curated records after reviewing the target collection and schema.
  • Delete entities with a reviewed filter expression.
  • Load or release collections while troubleshooting serving state.

Safety and Privacy

Milvus MCP is a database control surface. It can retrieve sensitive indexed content and can also mutate collections, indexes, entities, and database selection. Require human approval for writes, deletes, index creation, database switching, load, and release operations.

Treat vectors and metadata as sensitive even when source text is not returned. Embeddings, IDs, filters, collection names, and retrieval results can reveal private context. Restrict SSE and Streamable HTTP modes to trusted networks and avoid leaking Milvus tokens or query results into prompts, logs, or committed configuration.

Duplicate Check

Existing content includes a general Milvus tools entry in content/tools, but no Milvus MCP Server entry was found in content/mcp. This submission is scoped to zilliztech/mcp-server-milvus, the MCP server, and does not change the existing Milvus database listing.

Source citations

Add this badge to your README

Show that Milvus MCP Server is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.

Listed on HeyClaude
[![Listed on HeyClaude](https://heyclau.de/badge/mcp/milvus-mcp-server.svg)](https://heyclau.de/entry/mcp/milvus-mcp-server)

How it compares

Milvus MCP Server side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.

1 trust signal differ across this comparison (Submitter).

Field

MCP server from Zilliz for connecting Claude to Milvus vector database collections, text search, vector search, hybrid search, inserts, deletes, indexes, collection loading, database switching, and collection metadata.

Open dossier

Official Pinecone Developer MCP server that connects Claude and other MCP clients to Pinecone projects and documentation for index management, record upserts, semantic search, cascading multi-index search, reranking, and documentation lookup over integrated-inference indexes.

Open dossier

Connect Claude to a Weaviate vector database — run hybrid search, inspect collection config, list tenants, and upsert objects — using Weaviate's built-in Model Context Protocol server.

Open dossier

Official Chroma MCP server for connecting Claude to Chroma collections, documents, semantic search, full-text search, metadata filtering, persistent storage, self-hosted Chroma, and Chroma Cloud.

Open dossier
Next steps
Trust
Review statusNot reviewedNot reviewedNot reviewedNot reviewed
Package trustPackage not verifiedPackage not verifiedPackage not verifiedPackage not verified
Source provenanceSource-backedSource-backedSource-backedSource-backed
SubmitterDiffersoktofeesh1JSONboredoktofeesh1
Install riskReview firstReview firstReview firstReview first
Notes Safety ✓ Privacy ✓ Safety ✓ Privacy ✓ Safety ✓ Privacy ✓ Safety ✓ Privacy ✓
BrandMilvus logoMilvusPinecone logoPineconeChroma logoChroma
Categorymcpmcpmcpmcp
SourceSource-backedSource-backedSource-backedSource-backed
AuthorZillizPineconeWeaviateChroma
Added2026-06-062026-06-112026-06-172026-06-06
Platforms
Harness
Source repo
Safety notesMilvus MCP can read collection metadata, query collections, and run text, vector, text-similarity, multi-vector, and hybrid searches. Write-capable tools can create collections, insert data, upsert data, delete entities, create indexes, bulk insert records, load collections, release collections, and switch databases. The README notes that the environment file has higher priority than command-line arguments, so stale or unexpected environment settings can silently change the target Milvus instance. SSE and Streamable HTTP modes can expose database operations over HTTP and should be network-restricted. Remote Milvus or Zilliz Cloud credentials should be scoped to the collections and databases Claude is allowed to access.The server can create indexes and upsert records, so an agent with a write-capable API key can change live Pinecone project state. Run the npm package `@pinecone-database/mcp` through `npx`, which downloads and executes the published package on each launch; pin to a trusted version if reproducibility matters. Scope the Pinecone API key to the intended project and use read-only or least-privilege keys when index creation and writes are not needed. Require human review before `create-index-for-model` and `upsert-records` runs that mutate production indexes.The MCP server runs inside your Weaviate instance and respects its existing RBAC; scope the API key to least privilege. The object-upsert tool writes data — restrict write access to the collections Claude should modify.Chroma MCP can create, modify, and delete collections. Document tools can add, update, query, retrieve, and delete documents, metadata, custom IDs, and embeddings. Persistent, HTTP, and cloud modes can mutate durable retrieval stores rather than temporary test collections. External embedding functions can send document text or image-derived content to third-party embedding providers. Chroma Cloud and self-hosted HTTP modes require careful handling of tenants, databases, API keys, custom auth credentials, SSL settings, and network exposure. Retrieved context can influence Claude output even when stale, irrelevant, over-broad, or not authorized for the task.
Privacy notesMilvus collections can contain embeddings, sparse vectors, scalar fields, IDs, document chunks, metadata, image or multimodal references, query logs, and retrieval results that reveal sensitive project or user data. Milvus URI, tokens, database names, collection names, vector payloads, filter expressions, and retrieved records should stay out of prompts, issues, logs, screenshots, and committed files. Search results can include private source content that may be re-exposed in model transcripts or downstream tickets. HTTP transports, debug tools, query traces, failed-search artifacts, backups, and benchmark datasets need retention and access-control review.The `PINECONE_API_KEY` is read from the MCP client environment and grants access to the associated Pinecone project; keep it out of prompts, notes, and committed files. Index names, configurations, namespaces, statistics, record contents, and search queries can be exposed to the MCP client and model provider. Records and search text may contain embedded documents, customer data, or proprietary content, so review what is sent into indexes and returned by searches. Documentation search and tool calls reach Pinecone endpoints such as api.pinecone.io, so network access and request metadata leave the local machine.Query text, retrieved objects, and collection metadata enter the MCP client context and the model's prompt. The Weaviate endpoint URL and API key are secrets — keep them in the client config or environment, not in shared repositories.Collections can store source documents, chunks, embeddings, metadata, IDs, filters, and query results that reveal private project, customer, or research data. Query text, retrieved documents, metadata filters, and embedding inputs may expose sensitive information to Chroma Cloud, self-hosted operators, embedding providers, logs, or downstream model providers. CHROMA_API_KEY, custom auth credentials, embedding provider API keys, tenant IDs, database names, hostnames, and dotenv files should stay out of prompts, issues, logs, screenshots, and committed files. Persistent data directories and exported collections need the same access control, backup, encryption, and retention review as the source documents they index.
Prerequisites
  • Python 3.10 or newer.
  • uv installed for the README's recommended run path.
  • Running local or remote Milvus instance.
  • Milvus URI, token, and database name selected for the target environment.
  • Node.js 18 or newer with `npx` available.
  • Pinecone account and API key generated from the Pinecone console at app.pinecone.io.
  • MCP client that supports stdio server configuration, such as Claude Desktop or Cursor.
  • Awareness that index management and record tools require integrated-inference indexes.
  • A Weaviate instance on v1.37.1 or later (self-hosted or Weaviate Cloud).
  • The MCP server enabled on that instance via MCP_SERVER_ENABLED=true.
  • A Weaviate API key with the RBAC permissions for the collections Claude should reach.
  • An MCP client such as Claude Code or Claude Desktop.
  • Python 3.10 or newer available through uv for local stdio usage.
  • [object Object]
  • Storage, backup, and retention plan for persistent collections before enabling write tools.
  • Chroma Cloud tenant, database, and API key or self-hosted Chroma host credentials if using remote modes.
Install
git clone https://github.com/zilliztech/mcp-server-milvus.git && cd mcp-server-milvus && uv run src/mcp_server_milvus/server.py --milvus-uri <milvus-uri>
npx -y @pinecone-database/mcp
claude mcp add --transport http weaviate https://<your-weaviate-host>/v1/mcp
uvx chroma-mcp
Config
{
  "mcpServers": {
    "milvus": {
      "command": "uv",
      "args": [
        "--directory",
        "mcp-server-milvus",
        "run",
        "src/mcp_server_milvus/server.py",
        "--milvus-uri",
        "<milvus-uri>"
      ]
    }
  }
}
{
  "mcpServers": {
    "pinecone": {
      "command": "npx",
      "args": ["-y", "@pinecone-database/mcp"],
      "env": {
        "PINECONE_API_KEY": "{pinecone-api-key}"
      }
    }
  }
}
{
  "mcpServers": {
    "weaviate": {
      "url": "https://<your-weaviate-host>/v1/mcp",
      "type": "http"
    }
  }
}
{
  "mcpServers": {
    "chroma": {
      "command": "uvx",
      "args": ["chroma-mcp"]
    }
  }
}
Citations
ClaimUnclaimedUnclaimedUnclaimedUnclaimed
Open 4 picks in the interactive comparison tool

Signals

Loading live community signals…

More like this, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.