vector-mcp 1.11.1

Integrate RAG into AI Agents via MCP Server. Supports multiple Vector database technologies.

Vector Database - A2A | AG-UI | MCP

Version: 1.11.1

Overview

This is an MCP Server implementation which allows for a standardized collection management system across vector database technologies.

This was heavily inspired by the RAG implementation of Microsoft's Autogen V1 framework, however, this was changed to an MCP server model instead.

AI Agents can:

• Hybrid search for document information (lexical/vector)
• Create collections with documents stored on the local filesystem or URLs
• Add documents to a collection
• Utilize collection for retrieval augmented generation (RAG)
• Delete collection

Supports:

• ChromaDB
• PGVector
• Couchbase
• Qdrant
• MongoDB

This repository is actively maintained - Contributions and bug reports are welcome!

Automated tests are planned

MCP

MCP Tools

Function Name	Description	Tag(s)
create_collection	Creates a new collection or retrieves an existing one in the vector database.	collection_management
semantic_search	Retrieves and gathers related knowledge from the vector database instance using the question variable.	semantic_search
add_documents	Adds documents to an existing collection in the vector database. This can be used to extend collections with additional documents.	collection_management
delete_collection	Deletes a collection from the vector database.	collection_management
list_collections	Lists all collections in the vector database.	collection_management

A2A Agent

Architecture:

---
config:
  layout: dagre
---
flowchart TB
 subgraph subGraph0["Agent Capabilities"]
        C["Agent"]
        B["A2A Server - Uvicorn/FastAPI"]
        D["MCP Tools"]
        F["Agent Skills"]
  end
    C --> D & F
    A["User Query"] --> B
    B --> C
    D --> E["Platform API"]

     C:::agent
     B:::server
     A:::server
    classDef server fill:#f9f,stroke:#333
    classDef agent fill:#bbf,stroke:#333,stroke-width:2px
    style B stroke:#000000,fill:#FFD600
    style D stroke:#000000,fill:#BBDEFB
    style F fill:#BBDEFB
    style A fill:#C8E6C9
    style subGraph0 fill:#FFF9C4

Component Interaction Diagram

sequenceDiagram
    participant User
    participant Server as A2A Server
    participant Agent as Agent
    participant Skill as Agent Skills
    participant MCP as MCP Tools

    User->>Server: Send Query
    Server->>Agent: Invoke Agent
    Agent->>Skill: Analyze Skills Available
    Skill->>Agent: Provide Guidance on Next Steps
    Agent->>MCP: Invoke Tool
    MCP-->>Agent: Tool Response Returned
    Agent-->>Agent: Return Results Summarized
    Agent-->>Server: Final Response
    Server-->>User: Output

Usage

MCP CLI

Short Flag	Long Flag	Description
-h	--help	Display help information
-t	--transport	Transport method: 'stdio', 'http', or 'sse' [legacy] (default: stdio)
-s	--host	Host address for HTTP transport (default: 0.0.0.0)
-p	--port	Port number for HTTP transport (default: 8000)
	--auth-type	Authentication type: 'none', 'static', 'jwt', 'oauth-proxy', 'oidc-proxy', 'remote-oauth' (default: none)
	--token-jwks-uri	JWKS URI for JWT verification
	--token-issuer	Issuer for JWT verification
	--token-audience	Audience for JWT verification
	--oauth-upstream-auth-endpoint	Upstream authorization endpoint for OAuth Proxy
	--oauth-upstream-token-endpoint	Upstream token endpoint for OAuth Proxy
	--oauth-upstream-client-id	Upstream client ID for OAuth Proxy
	--oauth-upstream-client-secret	Upstream client secret for OAuth Proxy
	--oauth-base-url	Base URL for OAuth Proxy
	--oidc-config-url	OIDC configuration URL
	--oidc-client-id	OIDC client ID
	--oidc-client-secret	OIDC client secret
	--oidc-base-url	Base URL for OIDC Proxy
	--remote-auth-servers	Comma-separated list of authorization servers for Remote OAuth
	--remote-base-url	Base URL for Remote OAuth
	--allowed-client-redirect-uris	Comma-separated list of allowed client redirect URIs
	--eunomia-type	Eunomia authorization type: 'none', 'embedded', 'remote' (default: none)
	--eunomia-policy-file	Policy file for embedded Eunomia (default: mcp_policies.json)
	--eunomia-remote-url	URL for remote Eunomia server

Using as an MCP Server

The MCP Server can be run in two modes: stdio (for local testing) or http (for networked access). To start the server, use the following commands:

Run in stdio mode (default):

vector-mcp --transport "stdio"

Run in HTTP mode:

vector-mcp --transport "http"  --host "0.0.0.0"  --port "8000"

Creating Collection

AI Prompt:

Create a collection called zapdos with the documents in this directory:
/home/user/Documents/Chroma

AI Response:

The collection named "zapdos" has been successfully created or retrieved from the vector database
using the documents in the directory /home/user/Documents/Chroma.
Let me know if you'd like to perform any further actions, such as querying the collection or adding more documents!

Retrieving from Collection

AI Prompt:

Create a collection called zapdos with the documents in this directory:
/home/user/Documents/Chroma

AI Response:

The collection named "zapdos" has been successfully created or retrieved from the vector database
using the documents in the directory /home/user/Documents/Chroma.
Let me know if you'd like to perform any further actions, such as querying the collection or adding more documents!

Deleting Collection

AI Prompt:

Delete the collection called memory

AI Response:

The collection named "memory" has been successfully deleted.
Let me know if you'd like to create a new collection or perform any other actions!

A2A CLI

Endpoints

• Web UI: http://localhost:8000/ (if enabled)
• A2A: http://localhost:8000/a2a (Discovery: /a2a/.well-known/agent.json)
• AG-UI: http://localhost:8000/ag-ui (POST)

Short Flag	Long Flag	Description
-h	--help	Display help information
	--host	Host to bind the server to (default: 0.0.0.0)
	--port	Port to bind the server to (default: 9000)
	--reload	Enable auto-reload
	--provider	LLM Provider: 'openai', 'anthropic', 'google', 'huggingface'
	--model-id	LLM Model ID (default: nvidia/nemotron-3-super)
	--base-url	LLM Base URL (for OpenAI compatible providers)
	--api-key	LLM API Key
	--mcp-url	MCP Server URL (default: http://localhost:8000/mcp)
	--web	Enable Pydantic AI Web UI

Deploy MCP Server as a Service

The MCP server can be deployed using Docker, with configurable authentication, middleware, and Eunomia authorization.

Using Docker Run

docker pull knucklessg1/vector-mcp:latest

docker run -d \
  --name vector-mcp \
  -p 8004:8004 \
  -e HOST=0.0.0.0 \
  -e PORT=8004 \
  -e TRANSPORT=http \
  -e AUTH_TYPE=none \
  -e EUNOMIA_TYPE=none \
  knucklessg1/vector-mcp:latest

For advanced authentication (e.g., JWT, OAuth Proxy, OIDC Proxy, Remote OAuth) or Eunomia, add the relevant environment variables:

docker run -d \
  --name vector-mcp \
  -p 8004:8004 \
  -e HOST=0.0.0.0 \
  -e PORT=8004 \
  -e TRANSPORT=http \
  -e AUTH_TYPE=oidc-proxy \
  -e OIDC_CONFIG_URL=https://provider.com/.well-known/openid-configuration \
  -e OIDC_CLIENT_ID=your-client-id \
  -e OIDC_CLIENT_SECRET=your-client-secret \
  -e OIDC_BASE_URL=https://your-server.com \
  -e ALLOWED_CLIENT_REDIRECT_URIS=http://localhost:*,https://*.example.com/* \
  -e EUNOMIA_TYPE=embedded \
  -e EUNOMIA_POLICY_FILE=/app/mcp_policies.json \
  knucklessg1/vector-mcp:latest

Using Docker Compose

Create a docker-compose.yml file:

services:
  vector-mcp:
    image: knucklessg1/vector-mcp:latest
    environment:
      - HOST=0.0.0.0
      - PORT=8004
      - TRANSPORT=http
      - AUTH_TYPE=none
      - EUNOMIA_TYPE=none
    ports:
      - 8004:8004

For advanced setups with authentication and Eunomia:

services:
  vector-mcp:
    image: knucklessg1/vector-mcp:latest
    environment:
      - HOST=0.0.0.0
      - PORT=8004
      - TRANSPORT=http
      - AUTH_TYPE=oidc-proxy
      - OIDC_CONFIG_URL=https://provider.com/.well-known/openid-configuration
      - OIDC_CLIENT_ID=your-client-id
      - OIDC_CLIENT_SECRET=your-client-secret
      - OIDC_BASE_URL=https://your-server.com
      - ALLOWED_CLIENT_REDIRECT_URIS=http://localhost:*,https://*.example.com/*
      - EUNOMIA_TYPE=embedded
      - EUNOMIA_POLICY_FILE=/app/mcp_policies.json
    ports:
      - 8004:8004
    volumes:
      - ./mcp_policies.json:/app/mcp_policies.json

Run the service:

docker-compose up -d

Configure mcp.json for AI Integration

{
  "mcpServers": {
    "vector_mcp": {
      "command": "uv",
      "args": [
        "run",
        "--with",
        "vector-mcp",
        "vector-mcp"
      ],
      "env": {
        "DATABASE_TYPE": "chromadb",                   // Optional
        "COLLECTION_NAME": "memory",                   // Optional
        "DOCUMENT_DIRECTORY": "/home/user/Documents/"  // Optional
      },
      "timeout": 300000
    }
  }
}

Security & Governance

This project is built on agent-utilities, inheriting enterprise-grade security and governance features.

Authentication & Authorization

Feature	Description
OIDC Token Delegation	RFC 8693 token exchange for user-context propagation from A2A → MCP
Eunomia Policies	Fine-grained, policy-driven tool authorization (none, embedded, remote)
Scoped Credentials	Tools execute with the caller's scoped identity where possible
3LO / OAuth / API Token	Multiple auth strategies with graceful fallback

Eunomia Policy Enforcement

Eunomia provides a policy enforcement point for all tool calls:

• Embedded mode: Load local mcp_policies.json for role-based access, sensitivity gating, and audit logging
• Remote mode: Forward authorization decisions to a central Eunomia policy server for multi-agent governance
• Enable via CLI: --eunomia-type embedded --eunomia-policy-file mcp_policies.json

Runtime Protections

Protection	Description
Tool Guard	Sensitivity detection with human-in-the-loop approval gating
Prompt Injection Defense	Input scanning and repetition/loop guards
Content Filtering	Output schema enforcement and cost budget controls
Stuck Loop Detection	Automatic detection and recovery from agent loops
Context Limit Warnings	Proactive alerts before context window exhaustion

Graph Agent Architecture

The A2A agent uses pydantic-graph orchestration with:

• RouterNode: Lightweight classifier that routes queries to specialized domains
• DomainNode: Focused executor with only relevant tools loaded, preventing tool hallucination
• Approval Gates: Policy-driven approval workflows before sensitive operations
• Usage Guards: Budget and rate limiting enforcement

Production Recommendation: Enable --eunomia-type embedded (or remote) + OIDC delegation + containerized deployment. See agent-utilities documentation for full policy configuration.

Install Python Package

python -m pip install vector-mcp

PGVector dependencies

python -m pip install vector-mcp[postgres]

All

python -m pip install vector-mcp[all]

or

uv pip install --upgrade vector-mcp[all]

Repository Owners

Special shoutouts to Microsoft Autogen V1 ♥️

MCP Configuration Examples

stdio (recommended for local development)

{
  "mcpServers": {
    "vector": {
      "command": ".venv/bin/vector-mcp",
      "args": [],
      "env": {}
    }
  }
}

Streamable HTTP (recommended for production)

{
  "mcpServers": {
    "vector": {
      "url": "http://localhost:8080/vector-mcp/mcp"
    }
  }
}

Available MCP Tools

This server implements an action-routed dynamic tool architecture, consolidating operations into categorized tools.

Tool Name	Action	Description
vector_collection_management	create_collection	Executes create_collection within the collection_management category.
vector_collection_management	add_documents	Executes add_documents within the collection_management category.
vector_collection_management	delete_collection	Executes delete_collection within the collection_management category.
vector_collection_management	list_collections	Executes list_collections within the collection_management category.
vector_search	semantic_search	Executes semantic_search within the search category.
vector_search	lexical_search	Executes lexical_search within the search category.
vector_search	search	Executes search within the search category.