In this challenge, you will publish your agent so it can be called from external applications like Claude Desktop or Cursor.
Goal
Make your Northwind Analyst agent accessible via an MCP server endpoint, then connect to it from a host application.
Internal and External Access
Agents have two access modes:
Internal (default): Only members of your Aura project can use the agent. No additional charges.
External: The agent is exposed via REST API or MCP server. External agents incur charges per Neo4j pricing.
For this challenge, you need External access with MCP server enabled.
Steps
-
Open your agent in the Aura Console
-
Click the agent menu and select Configure
-
Under Access, select External
-
Enable the MCP server toggle
-
Click Save
-
From the agent menu, click Copy MCP server endpoint
-
Add the endpoint to your host application (see below)
-
Send a test message to confirm the agent responds
Connecting to Claude Desktop
-
Open Claude Desktop settings or configuration file
-
Add your agent’s MCP endpoint
-
Configure authentication if required
-
Restart Claude Desktop
-
Send a test question to your agent
See Aura Agent documentation for detailed Claude Desktop setup instructions.
Connecting to Cursor
-
Open Cursor settings → MCP configuration
-
Add a new MCP server with your agent’s endpoint
-
Save and reload Cursor
-
Use the agent through Cursor’s interface
Summary
In this challenge, you published your agent as an external MCP server and connected to it from a host application.
Next Steps
Continue with the GraphRAG track:
-
Neo4j GraphRAG for Python - Build knowledge graphs using the neo4j-graphrag package
-
Using Neo4j with LangChain - Integrate Neo4j with LangChain for RAG and agents
-
New Cypher AI Procedures - GraphRAG in pure Cypher with
ai.text.embed,ai.text.embedBatch, andai.text.completion