Key Takeaways
- Connect Claude AI to Oracle EPM Cloud using MCP and official REST APIs
- Run business rules, export data, check variables — through natural language
- Open-source MCP server available on GitHub
- Custom integration — not an Oracle product
- Same security model as EPM Cloud UI — no elevated permissions
What Is This?
This is a custom MCP (Model Context Protocol) server that connects Claude AI directly to Oracle EPM Cloud. MCP is an open standard created by Anthropic that allows AI models to interact with external systems through a standardized interface. Think of it as a bridge between an AI model and the systems it needs to interact with.
For Oracle EPM Cloud, this means you can talk to Claude in plain English and have it make REST API calls to your EPM environment on your behalf. Instead of logging into the workspace, writing EPM Automate scripts, or building Postman collections — you just ask.
This is not an Oracle product. Oracle has built MCP servers for Oracle Database and Autonomous Database, but not for EPM Cloud. This is a custom integration built using Oracle's official, documented REST APIs — the same APIs that EPM Automate uses behind the scenes.
The Architecture
↓
CLAUDE DESKTOP (AI Assistant)
↓ MCP Protocol
MCP SERVER (Custom — open source on GitHub)
↓ REST API Calls
ORACLE EPM REST APIs (Official, documented)
↓ HTTPS / JSON
YOUR EPM CLOUD (Planning · FCCS · PCM)
You talk to Claude Desktop on your computer. Claude determines which tool to use based on what you asked. The MCP server translates that into the appropriate REST API call. The response comes back to Claude, who explains the results in plain English.
How It Works
Each tool in the MCP server maps directly to a documented Oracle EPM REST API endpoint. For example, you can say "Show me the current substitution variables for Vision" and Claude calls the appropriate tool, which hits the EPM Planning REST API endpoint. You get back CurrMonth, CurrYear, CurrScenario — all the values that control what your finance team sees in their forms and reports.
The server covers a complete month-end close workflow: check the current period, run the aggregation, wait for completion, validate data, and roll the period forward. All through natural language.
How to Set It Up
You need two things: Node.js 18+ and Claude Desktop.
Clone the repository from GitHub, run npm install, and configure Claude Desktop to point to the MCP server. The server runs in mock mode by default — you don't need an EPM environment to test the setup. When you're ready to connect to a real environment, you update one config file.
Full setup instructions are in the GitHub README. The GitHub README explains the structure.
What's Next
Part 2 will connect to a live Oracle EPM environment with real data and real business rules. Future plans include OAuth 2.0 support for MFA environments and FCCS-specific tools.
The server is open source (MIT license). Add tools by defining them in index.js.
Related Articles
Automating Clone of Production to Test with Groovy
Form Performance Testing with simulateConcurrentUsage
More from FMEPM
Get Oracle EPM tutorials, walkthroughs, and practical project lessons.
Working with Oracle EPM?
Let's talk about how AI can fit into your EPM environment.
Book a Discovery Call →Resources
GitHub Repository: github.com/fmepm/oracle-epm-mcp-server
Oracle EPM REST API Docs: Oracle Documentation
Oracle EPM 26.04 Checklist: What Changed and What To Do
FMEPM · Fred Mamadjanov · Oracle ACE · EPM Solution Architect
This is not an Oracle product. Oracle EPM Cloud is a trademark of Oracle Corporation. This page describes a custom integration using Oracle's official, documented REST APIs.