Knowledge Graphs for AI Agent API Discovery
by DeepLearning.AI and SAP
Build a knowledge graph of your APIs and business processes so an agent finds the right endpoints and calls them in the right order.
Overview
Knowledge Graphs for AI Agent API Discovery is an intermediate DeepLearning.AI short course made with SAP Business AI and taught by Pavithra G K, head of business knowledge graphs at SAP Business AI, and Lars Heling, a senior knowledge engineer there. The course page lists it as updated July 17, 2026. It addresses a problem that plain tool calling handles badly: an organization with thousands of APIs, where an agent has to pick the right service and also call prerequisite APIs in the order a business process requires. The seven lessons run about 74 minutes. After an introduction to knowledge graphs for agents, the code lessons build an API knowledge graph from raw API specifications, integrate business process data to link APIs that were previously isolated, and implement discovery that starts with semantic retrieval and then follows process connections to find missing prerequisite APIs and their calling order. The final lesson builds a business process agent with three tools (discovery, GET data and POST data) that retrieves API metadata such as properties, navigations and process information before acting. Heling and Christoph Meyer presented the same approach at DeepLearning.AI's AI Dev 25 NYC event. There they explained that SAP's graph combines OpenAPI specifications with process models such as BPMN, embeds each node together with its neighbors, and needs maintenance and subgraphs as it grows. The page does not name the graph database or LLM used.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Freemium
- Duration
- ~1h 14m of video and code, plus a 9-minute graded quiz, self-paced
- Provider
- DeepLearning.AI and SAP
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Basic knowledge graph concepts and why agents need them for discovering services in large API ecosystems
- ✓Build a knowledge graph from raw API specifications and visualize its services and endpoints
- ✓Extend the graph with business process data to link previously isolated APIs to each other
- ✓Combine semantic retrieval with process connections to find prerequisite APIs and their correct calling order
- ✓Build an agent with discovery, GET data and POST data tools that executes multi-step API tasks
- ✓Retrieve API metadata such as properties, navigations and process information to ground each agent action
Highlights
- •Taught by SAP Business AI knowledge engineers who build API knowledge graphs for SAP's own API catalog
- •Covers the call-order problem (prerequisite APIs from business processes), which most tool-calling tutorials ignore
- •Uses one small set of generic tools instead of one tool per API, which works when an agent faces thousands of endpoints
- •Hands-on: four code lessons that build from graph construction to a working business process agent
Who It’s For
Best For
- ✓Developers building agents that work with large enterprise API catalogs such as ERP or SaaS platforms
- ✓RAG engineers who want to add graph structure on top of vector search for tool retrieval
- ✓Platform teams designing API discovery or MCP tool registries for internal agents
Prerequisites
- •Basic Python knowledge, as stated on the course page
- •Familiarity with REST APIs and API specifications such as OpenAPI helps with the graph-building lessons
- •Some exposure to embeddings or semantic search makes the discovery lesson easier to follow
FAQ
What is Knowledge Graphs for AI Agent API Discovery?
A DeepLearning.AI short course with SAP Business AI for developers whose agents must work across large API estates. You build a knowledge graph from API specifications, link it to business process data, and finish with an agent that discovers, orders and executes the API calls a real task needs.
Is Knowledge Graphs for AI Agent API Discovery free?
Knowledge Graphs for AI Agent API Discovery offers free content, with paid options for certificates or premium features.
What level is Knowledge Graphs for AI Agent API Discovery for?
Knowledge Graphs for AI Agent API Discovery is aimed at a intermediate audience. Recommended background: Basic Python knowledge, as stated on the course page, Familiarity with REST APIs and API specifications such as OpenAPI helps with the graph-building lessons, Some exposure to embeddings or semantic search makes the discovery lesson easier to follow.
How long does Knowledge Graphs for AI Agent API Discovery take?
Expect roughly ~1h 14m of video and code, plus a 9-minute graded quiz, self-paced. Most learners work through it at their own pace.
What will I learn from Knowledge Graphs for AI Agent API Discovery?
You'll learn: Basic knowledge graph concepts and why agents need them for discovering services in large API ecosystems; Build a knowledge graph from raw API specifications and visualize its services and endpoints; Extend the graph with business process data to link previously isolated APIs to each other; Combine semantic retrieval with process connections to find prerequisite APIs and their correct calling order; Build an agent with discovery, GET data and POST data tools that executes multi-step API tasks; Retrieve API metadata such as properties, navigations and process information to ground each agent action.
Topics
Sources
This page was written from 3 sources, 1 on domains other than deeplearning.ai.