AnythingLLM

You can connect AnythingLLM to MCP-compatible clients using an MCP server that exposes AnythingLLM capabilities through simple, client-friendly endpoints. This lets you manage workspaces, chat, documents, embeddings, users, API keys, and more from tools built for MCP-enabled clients like TypingMind and Claude Desktop.

How to use

You will run an MCP server that interfaces with your AnythingLLM instance and then configure your MCP client to load that server. Once linked, you can manage workspaces, chat with AI, upload documents, embed content, search across your data, handle users and API keys, and control system settings all from your MCP client. Start by adding the AnythingLLM MCP server in your client and initializing it with your AnythingLLM credentials. After initialization, you can perform common tasks through natural language prompts or dedicated tools provided by the client.

Configuration snippets below show two explicit MCP server entries you can use. One is for TypingMind and the other for Claude Desktop. Each entry runs via npx to launch the AnythingLLM MCP server package.

How to install

Prerequisites you need before starting:

• Node.js 18+ installed

• AnythingLLM instance running with API access enabled

• AnythingLLM API key (obtain from your AnythingLLM instance at /settings/api-keys)

Choose one of the installation options below to run the MCP server.

Additional configuration and usage notes

Configure your MCP client to connect to the AnythingLLM MCP server using the provided entries. The server entries below show how to start the MCP server with npx and the appropriate package reference.

{
  "mcpServers": {
    "anythingllm_typingmind": {
      "command": "npx",
      "args": ["-y", "anythingllm-mcp-server@2.0.0"]
     }
   }
}

{
  "mcpServers": {
    "anythingllm_claude": {
      "command": "npx",
      "args": ["-y", "anythingllm-mcp-server@1.0.3"]
    }
  }
}

Initialization

When you first connect, initialize the AnythingLLM MCP client with your API credentials from your AnythingLLM instance. Use the initialize_anythingllm tool with apiKey and baseUrl parameters as shown in your MCP client’s init flow.

Security and maintenance

• Keep your API key secure and never commit it to version control.

• Use environment variables for sensitive configuration in production.

• Restrict API key permissions to only what is needed.

Troubleshooting

• Ensure your AnythingLLM instance is running and accessible and that the base URL includes the protocol (http or https). Verify the API key has proper permissions.

• If you see initialization or connection errors, re-run the initialize_anythingllm tool with your apiKey and baseUrl.

• Check workspace slugs and verify connectivity to the AnythingLLM API if you encounter "Workspace not found" or similar messages.

Notes

This MCP server exposes a wide range of capabilities from AnythingLLM, including workspace and chat management, document handling, embedding and search, user and API key administration, system settings, and agent support. You can mix and match client configurations for TypingMind, Claude Desktop, or other MCP-enabled clients.

Available tools

initialize_anythingllm

Initialize the AnythingLLM client with API credentials (apiKey and baseUrl).

list_workspaces

Get all workspaces within the connected AnythingLLM instance.

get_workspace

Retrieve details for a specific workspace by slug.

create_workspace

Create a new workspace with a given name.

update_workspace

Update settings for a specific workspace.

delete_workspace

Delete a workspace by slug.

get_workspace_settings

Get current settings for a specific workspace.

update_workspace_settings

Update settings for a specific workspace.

chat_with_workspace

Send a message to a workspace and receive responses.

get_chat_history

Fetch chat history for a workspace.

clear_chat_history

Clear all chat history for a workspace.

list_documents

List all documents in a workspace.

delete_document

Remove a document from a workspace.

process_document_url

Process and ingest a document from a URL.

get_document_vectors

Get vector embeddings for a document.

embed_text

Embed raw text into a workspace.

embed_webpage

Embed a webpage into a workspace.

search_workspace

Search content within a workspace.

list_users

List all users in the system.

create_user

Create a new user with a username and password.

update_user

Update an existing user.

delete_user

Delete a user by userId.

list_api_keys

List all API keys.

create_api_key

Create a new API key.

delete_api_key

Delete an API key by keyId.

get_system_settings

View system configuration.

update_system_settings

Modify system settings.

get_system_info

Get general system information.

get_system_stats

Get system statistics.

list_llm_providers

List available LLM providers.

update_llm_provider

Update LLM provider configuration (provider, apiKey, model).

get_vector_database_info

Get vector database configuration.

update_vector_database

Update vector database configuration (provider, config).

list_agents

List all available agents.

create_agent

Create a new agent with a systemPrompt and tools.

update_agent

Update an existing agent.

delete_agent

Delete an agent by agentId.

invoke_agent

Invoke an agent with input text.