Prerequisites
- AnythingLLM Desktop or self-hosted — anythingllm.com
- Node.js 18+
- Agent MCP Studio at agentmcp.studio
Step-by-Step Setup
Get your relay URL
Open the studio → Settings → MCP Relay Bridge. Copy your session URL and click Connect.
Download bridge.js
curl -O https://agentmcp.studio/bridge.js && npm install ws
-
Add MCP server in AnythingLLM
In AnythingLLM: Settings → Agent Skills → MCP Servers → Add MCP Server:
{ "name": "agent-mcp-studio", "command": "node", "args": ["/path/to/bridge.js", "wss://agentmcp.studio/api/relay/YOUR-UUID"] }Save and toggle the server On. AnythingLLM will show your tools in the agent's skill list.
Use tools in an agent workspace
Open a workspace, switch to Agent mode (the robot icon), and start a conversation. The agent will automatically call your tools when relevant.
Frequently Asked Questions
Yes — AnythingLLM supports MCP via its Agent Skills system. Configure MCP servers in Settings → Agent Skills → Custom Skills. Your Agent MCP Studio tools appear as callable skills in AnythingLLM's agent mode.
AnythingLLM is a popular open-source, self-hosted AI assistant that works with any LLM (OpenAI, Anthropic, local Ollama models). It supports RAG, multi-user workspaces, and extensible agent skills via MCP.
Yes — AnythingLLM works with any LLM that supports function calling, including Ollama-hosted models like Llama 3, Mistral, and Qwen. MCP tools from Agent MCP Studio are available regardless of the backend model.
In AnythingLLM, enable Agent Mode in your workspace settings (the toggle in the workspace config). Once enabled, the chat interface shows an agent icon and your MCP tools become available to the agent.
Yes — MCP server configuration in AnythingLLM is global. Tools configured in Settings → Agent Skills are available to all workspaces that have agent mode enabled.