The Fifth Elephant 2025 Annual Conference CfP

Abstract

As we begin to rely more on AI agents to assist us, it is also important to realize that they are only as useful as the knowledge they have access to.

A substantial amount of modern development relies on external integrations, and developers make heavy use of documentation, ranging from precise technical API specifications to verbose, text-heavy tutorials. With a wide array of integrations and an unpredictable pace of changes, developers have to constantly track changes and incorporate them into their work.

Traditional document management systems will often fall short in keeping pace with the evolving APIs, their specs, and developer guides. It goes without saying that we should be leveraging AI agents to make the lives of developers easier here. But, agents are not magical entities. They perform only as well as the instructions and context they are fed. And that is our responsibility.

In this talk, we present an approach that combines:

• Intelligent web crawling
• Semantic search
• Model Context Protocol (MCP)
• Change Impact Analysis and Context Structuring: extract structured, actionable context for AI agents from documentation diffs

to create a precise, well-structured system that agents can use to help developers keep their code integrations up-to-date with changes in third-party systems. This system provides the ability to detect and act on API contract changes, such as deprecations and schema modifications that introduce new, salient features. Instead of merely flagging these changes, the system can automatically contextualize them and highlight the exact places that need updates. This granular, versioned information directly feeds into AI workflows, enabling timely, proactive adjustments to code and logic.

The session will include demonstrations of real-time documentation monitoring and AI agent adaptation to evolving APIs, offering concrete examples of the approach in practice. Attendees will leave with actionable insights into designing resilient, adaptive AI systems capable of handling real-world complexity.

Key Takeaways

• The challenges of documentation drift and its impact on AI system reliability
• Designing and implementing self-updating knowledge bases using intelligent crawling, semantic search, and versioned change tracking
• Leveraging Model Context Protocol (MCP) for precise, contextualized change detection in APIs and documentation
• Practical demonstrations of AI systems responding to real-time changes, with clear visibility into what has changed and how
• Best practices for building robust, self-adapting AI workflows in production environments

Target Audience

• Engineering teams using AI-assisted development tools
• Platform Engineers managing multiple API integrations
• Development leaders seeking to eliminate documentation drift as a source of technical debt.

Open Source Link: https://github.com/iamakash-06/Doc-Monitor-MCP
Reference Link: https://sahaj.ai/mcp-and-doc-monitor-transform-external-services-integration/
Draft Slides: https://docs.google.com/presentation/d/1cpI83sa7rfvJBs50dS00GBQH1uvbUiUm9-Vs7s2fqfE/edit?usp=sharing

Speaker Bio

Akash Sathish is a Solution Consultant at Sahaj Software and a machine learning enthusiast with hands-on experience in AI-assisted solution delivery. He has delivered talks on Machine Learning in collaboration with DeepLearning.AI and Intel. Known for his knack for storytelling, Akash enjoys breaking down complex concepts into simple, relatable insights and making advanced topics accessible to a wide range of audiences.

Profile Links and Previous Talk links:

1. LinkedIn
2. Privacy Preserving Machine Learning - Sahaj DevDay
3. Breaking Into AIoT - Intel IoT Club in Collaboration with Deep Learning.AI
4. Breaking Into AI - Intel IoT Club