Generate agent and MCP Server pipelines

Use SnapGPT to generate agent pipelines and MCP Server pipelines for AI agent integration from natural language prompts.

In addition to standard pipelines, SnapGPT can generate two specialized pipeline types for AI agent integration:

  • Agent pipelines: A pipeline bundle in which a driver pipeline runs an LLM-based agent that calls one or more tool pipelines as callable tools. Refer to Build your first agent pipeline.
  • MCP Server pipelines: A pipeline bundle that exposes SnapLogic pipelines as tools through the Model Context Protocol (MCP), so that external AI clients can discover and invoke them. Refer to MCP Server Pipeline Builder.

Both types are generated using natural language prompts. No dedicated skill button or UI option is required. Describe the pipeline you want and SnapGPT generates the appropriate bundle.

Generate an agent pipeline

An agent pipeline consists of two pipelines generated as a bundle:

  • A driver pipeline (prefixed AGENT_): Orchestrates the agent loop using two Prompt Generator Snaps (one for the system prompt, one for the user prompt), a Multi Pipeline Function Generator Snap, and an OpenAI Chat Completions Agent Snap.
  • A tool pipeline (prefixed TOOL_): Performs the integration work that the driver pipeline registers and calls as a tool.
Important: You must attach an OpenAI account in Designer before executing the generated agent pipeline.

Example

In SnapGPT, enter a prompt such as:

Generate an agent pipeline that gets the current weather

SnapGPT describes the driver and tool pipelines it will build, including the Snaps used and any accounts required:


SnapGPT response describing the AGENT_Weather driver pipeline and TOOL_GetCurrentWeather tool pipeline it will generate

After generation, SnapGPT presents both pipelines as a bundle with a View details link for each pipeline. Click Import All to add both pipelines to your project.


Agent pipeline bundle showing the AGENT_Weather driver pipeline and the TOOL_GetCurrentWeather tool pipeline with the Import All button

Generate an MCP Server pipeline for AI agent integration

An MCP Server pipeline consists of two pipelines generated as a bundle:

  • An MCP Server pipeline (prefixed MCP_, tagged mcp_server): Handles MCP protocol routing using a Multi Pipeline Function Generator, MCP Server Router, Pipeline Execute, Union, and Remove Original.
  • A tool pipeline (prefixed TOOL_, tagged tool): Performs the integration work that the MCP Server exposes as a callable tool to AI clients.
Important: SnapGPT generates the MCP Server pipeline and the associated tool pipeline. The MCP Server asset and policies are not generated. After importing the pipelines, refer to MCP Server Setup to complete the deployment.

Example

In SnapGPT, enter a prompt such as:

Generate an MCP Server pipeline that gets the current exchange rate

SnapGPT describes the MCP Server pipeline and tool pipeline it will build, including the Snaps used and the API it will call:


SnapGPT response describing the MCP_ExchangeRate_Server pipeline and TOOL_GetExchangeRate tool pipeline it will generate

After generation, SnapGPT presents both pipelines as a bundle with a View details link for each pipeline. Click Import All to add both pipelines to your project.


MCP Server pipeline bundle showing the MCP_ExchangeRate_Server pipeline and the TOOL_GetExchangeRate tool pipeline with the Import All button

Related information

  • For background on the Model Context Protocol in SnapLogic, refer to Model Context Protocol (MCP).
  • To complete MCP Server deployment after generating the pipeline (create the MCP Server asset, configure authentication, and add policies), refer to MCP Server Setup.
  • To generate an MCP Server pipeline using the UI wizard instead of SnapGPT prompts, refer to MCP Server Pipeline Builder.