FME support for MCP

Safe Software has turned #FME into an #MCP bridge (to the #Esri #ArcGIS platform, for example): FME workspaces can now be published as AI-ready tools and served to any MCP client via FME Flow. Safe Software frames this as turning GIS platforms into something #AI agents can operate conversationally, with notably pitching FME as a “data guardian” shielding raw data from the AI clients that call it.
Author

Ralph Straumann

Published

July 9, 2026

MCP1 capabilities are now available “bi-directionally” in FME2. The MCPCaller Transformer3 is available from FME 2026.1 onwards, and FME Flow4 can act as an MCP Server since FME 2026.2.

The latter scenario is described in detail in a recent article by Safe Software. Basically, the new MCP capability means you can author a data workflow in FME Form5 and register it as an AI- or agent-ready tool with an MCP server run by FME Flow. Once published and registered, the tool (your workspace) can be called from any MCP client.

MCP architecture based on FME Flow’s MCP Server

The article gives the following account of the workflow:

  • Prepare and publish: Replace the workspace’s terminal step (…) with an MCP writer, which defines what comes back to the client: text, structured content (JSON6), images, resources, or embedded content. Adding it flags the workspace as an MCP tool. (…) Return only what the client needs (…). Keep input parameter types simple (text, numbers, booleans, dates, dropdowns) (…).

  • Register the tool: A single [FME] Flow instance can host multiple MCP servers, each with multiple tools, organized by domain or user group. (…) Add a tool pointing at your published workspace, then describe each parameter (…). The pattern throughout is: define the field, give the constraints, and say when the client needs more information.

  • Connect a client and test it out: Most AI clients let you add an MCP server by URL. A deliberately messy prompt is a good stress test (…) Underneath, [the task] is still a standard [FME] Flow job, visible in job history with its inputs and processing details; the only difference is that MCP triggered it.

In the article, Safe Software goes into more detail about the workflow and lists some caveats and principles for designing good MCP tools. It also describes a more advanced scenario involving an AI planner-in-a-loop that can invoke multiple tools in a single workflow and combine their results.

While Safe Software underlines that the principles shown in the article apply to any integration, not just ArcGIS-based ones, I found the statement that this turns ArcGIS into a recipient for conversational agents a particularly interesting notion: “(…) a small set of focused tools (…) turns ArcGIS into something an agent can operate conversationally.” Meanwhile, Esri seems to be working on similar ideas and has published an Early Adopter site on MCP in GIS which states:

Esri is working towards exposing GIS capabilities via Model Context Protocol (MCP) across ArcGIS Enterprise, Online and Location Platform7. Our intent with this work is to help you access GIS data and functionality through AI applications.

Finally, the Safe Software article extols information security as a key positive aspect8 of the presented usage scenario:

You can also control exactly what data is exposed to AI in any deployment using FME as a Data Guardian. The MCP clients never get direct access to your raw data, and in fact have no idea what the raw data looks like.

If MCP within the FME ecosystem interests you, Safe Software has much more MCP-related content on a dedicated “Solutions” page and offers (past and upcoming) webinars on the topic.

Footnotes

  1. Model Context Protocol, an open standard (introduced by Anthropic in late 2024) that lets AI applications connect to external tools, data sources, and services through a common interface.↩︎

  2. Safe Software’s Feature Manipulation Engine.↩︎

  3. The MCPCaller serves as an MCP (Model Context Protocol) client that can connect to an MCP server to call tools to work with resources and prompts.↩︎

  4. FME Flow is Safe Software’s server product for automating, scheduling, and hosting FME workspaces, formerly known as FME Server.↩︎

  5. FME Form is Safe Software’s desktop application for authoring FME workspaces, formerly known as FME Desktop or FME Workbench.↩︎

  6. JavaScript Object Notation.↩︎

  7. MCP support for the ArcGIS Location Platform is currently in beta.↩︎

  8. Note the capitalisation in the quote.↩︎