Make your own event-sourced agent harness using stream processors — Jonas Templestein, Iterate
TL;DR
Jonas Templestein and Misha from Iterate demonstrate a prototype event-sourced architecture for building distributed AI agent harnesses where all state changes are captured as immutable events in HTTP-accessible streams, enabling debuggability and composability across different languages and environments.
💡 Event-Sourced Architecture Philosophy 2 insights
Immutable event logs enable perfect debugging
Every state change is captured as an event with a type and payload, creating a complete audit trail that makes agent behavior fully reproducible and transparent.
Single abstraction simplifies system design
The architecture treats everything—from LLM responses to errors and scheduled tasks—as uniform events, eliminating complex side effects and special cases found in traditional agent frameworks.
🌐 Distributed & Edge-First Design 2 insights
Publicly routable HTTP endpoints for every agent
Agents exist as internet-connected server programs with unique URLs from creation, eliminating the need for connector frameworks to integrate external services like Slack or web forms.
Cross-language composability
Plugins and extensions can run on different machines written in different languages (Rust, TypeScript) and communicate purely through the event stream interface.
⚙️ Technical Implementation & Safety 3 insights
Circuit breaker prevents infinite loops
Streams automatically pause when exceeding 100 events per second, requiring manual intervention to resume and preventing runaway feedback loops between distributed components.
Native scheduling and webhook support
The system supports scheduled events (heartbeats, future tasks) and bidirectional HTTP subscriptions via Server-Sent Events, allowing agents to both consume and push events to external endpoints.
Tolerance for unstructured data
The API accepts arbitrary webhook payloads from third parties without strict validation, relying on LLMs to parse unstructured data rather than enforcing rigid schemas.
Bottom Line
Build AI agent harnesses as distributed, event-sourced systems with immutable logs and HTTP-native interfaces to maximize debuggability, composability, and interoperability across different languages and environments.
More from AI Engineer
View all
Frontier results, on device - RL Nabors, Arize
Rachel Lee Neighbors introduces a framework for replacing expensive cloud-based frontier models with Small Language Models (SLMs) running on-device, demonstrating how a systematic 'prototype big, deploy small' approach using evaluation tools like Phoenix can cut inference costs to zero while maintaining 90% accuracy and enabling offline functionality.
The Future Is Domain-Specific Agents - Justin Schroeder, StandardAgents
Justin Schroeder argues that the future of AI lies in domain-specific agents—small, specialized agents that compose together rather than general-purpose agents bloated with tools and skills, delivering 80%+ token efficiency and 137x cost savings compared to monolithic approaches.
The Agentic AI Engineer - Benedikt Sanftl, Mutagent
Benedikt Sanftl and Burak from Mutagent present the 'Agentic AI Engineer' paradigm, where specialized AI agents autonomously manage the entire lifecycle of building, evaluating, and optimizing other agents through automated offline and online loops, solving the scalability bottlenecks of manual development.
Bypassing the Multimodal Tax: Hybrid RAG, SQL RRF & UI Telemetry - Abed Matini, Ogilvy
Abed Matini presents a framework-free Hybrid RAG architecture that eliminates pre-query token costs by preprocessing documents locally using DocLink and multiple chunking strategies, while implementing SQL-based Reciprocal Rank Fusion and LangFuse telemetry for production observability.