Creating and Managing AI Toolkit Services

pgEdge Cloud databases can be deployed with an installed and configured MCP server, ready for connections. After deployment, use the Services dialog to open the Add MCP Server popup to add AI functionality to an existing cluster or to manage defined functionality.

Adding an MCP Server

Select the + Add MCP Server button to access the Add MCP Server popup to define an MCP server and optionally enable an LLM.

Use the fields on the Add MCP Server popup to describe the server and, optionally, the LLM:

• Use the Select Host field to select the cluster host on which this MCP server will be provisioned and run.

• Use the Target Nodes field to optionally select the database nodes this MCP server connects to, in priority order. Defaults to all nodes.

• Use the Allow Writes? toggle to optionally grant the MCP service read-write access (INSERT / UPDATE / DELETE) via the query_database tool. Note that allowing read-write access could potentially expose data to unexpected or unwanted modifications.

• Use the LLM Enabled? toggle to optionally enable an LLM to generate embeddings for the database. When the toggle is on, Cloud activates the generate_embedding tool on the MCP server and requests LLM provider credentials. To enable an LLM, provide the following information:

  • Use the Embedding Provider field to select the provider used by the generate_embedding tool on the MCP server.

  • Use the Embedding Model field to specify the model identifier used by the generate_embedding tool (e.g. text-embedding-3-small, voyage-3).

  • Use the Embedding API Key field to enter the API key for the selected embedding provider. This key is stored encrypted server-side.

When you've defined the MCP server (and optionally LLM functionality), select the + Add MCP Server button to update the database. The Services dialog displays the MCP server deployment details when the deployment is complete.

To delete an MCP Server, use the menu in the upper-right corner of the MCP Servers details panel; select Delete Service to access a confirmation popup that prompts you to enter the database name as confirmation that you wish to delete the service.

Connecting a Client to the MCP Server

The steps for connecting a client to the MCP server vary by client and platform. The Services dialog displays connection details for several popular clients under the Connect to MCP Clients label:

Select a tab to view and copy connection details for the selected client. Choose from:

• Claude Code
• Cursor
• OpenAI Codex
• Replit

Adding a RAG Server

Select the Add RAG Server button to access the Add RAG Server popup to define a RAG server and optionally enable an associated LLM.

Use the fields on the Add RAG Server popup to describe the server:

• Use the Select Host field to select the cluster host on which this RAG server service will be provisioned and run.

• Use the Default Token Budget field to set the maximum number of context tokens (500–128,000) the LLM can process per request.

• Use the Default Top N field to set the number of results retrieved from the vector store before they are trimmed to fit within the token budget.

• Use the Default Embedding LLM Provider field to select the provider whose model will generate vector embeddings for queries and documents during retrieval.

• Use the Default Embedding LLM Model field to specify the embedding model to use. This must match the model used to generate any pre-existing embeddings in the dataset.

• Use the Default Embedding LLM API Key field to enter the API key for authenticating with the selected embedding provider.

• Use the Default Completion LLM Provider field to select the provider whose model will be used for answer generation after relevant documents are retrieved.

• Use the Default Completion LLM Model field to specify the model used for answer generation. Select a suggested model or enter a custom one.

• Use the Default Completion LLM API Key field to enter the API key for authenticating with the selected completion provider.

Expand the Add Pipelines field to define one or more named pipelines. Each pipeline specifies one or more tables and their associated columns and vector columns.

Use the:

• + Add Table button to access fields to define additional tables. Then:

  • Use the Table Name field to specify the fully qualified name of the table to use for the pipeline (like public.documents).

  • Use the Text Column field to specify the column that contains the text content to be indexed and searched (like content).

  • Use the Vector Column field to specify the column that stores the vector embedding used for similarity search (like embedding).

• + Add Pipeline button to define additional pipelines and associated tables.

When you're finished defining the RAG server, select the + Add RAG Server button. The Services dialog displays the RAG server deployment details when the deployment is complete.

To delete a RAG Server, use the menu in the upper-right corner of the RAG Servers details panel; select Delete Service to access a confirmation popup that prompts you to enter the database name as confirmation that you wish to delete the service.

Using the RAG Server

After adding a RAG Server to your cloud deployment, you can use the pgEdge Docloader to load your documents into your database. The Docloader converts HTML, Markdown, and reStructuredText into a documents table:

pgedge-docloader --config docloader.yml

Once data is loaded, you can query your pipeline via the REST API. For example:

curl -X POST https://<your-rag-server-url>/v1/pipelines/my-docs/search \
  -H "Content-Type: application/json" \
  -d '{"query": "How do I configure replication?"}'

The RAG server retrieves the most relevant document chunks using hybrid search (vector similarity + BM25 keyword matching), then passes them to the LLM to generate a grounded answer.