Choosing agent architectures: stateful graphs, durable workflows, queues and supervisors
goyalshaliniuk · x · 2026-11-09
A practical architecture selection guide for long-running AI agents, mapping five patterns to use cases:
- Stateful graphs: multi-step workflows with complex state
- Durable workflows: tasks that must survive failures and interruptions
- Event-driven systems: agents reacting to external changes
- Supervisor–worker: complex goals requiring specialized agents
- Queue-based distributed architecture: high-volume async workloads; durable message queues, independent workers, retries/dead-letter queues, workload-based scaling
Key takeaway: long-running agents are persistent software systems needing memory, execution control, recovery, and reliable coordination — the real challenge is reliable long-term operation, not just staying active longer.
Related event: Five Architecture Patterns for Long-Running AI Agents(4 posts)→
More from coding & agent
- Multica teases customizable themes in next release — jiayuan_jy · 2026-11-09
- 'AI code is a mess but no-one can remember what good code looks like anymore' — kylegawley · 2026-11-09
- Queue-Based Distributed Architecture: Scale Agents with Celery-Style Task Queues — goyalshaliniuk · 2026-11-09
- Event-Driven Architecture: Trigger Agents from Kafka-Style Events, Not Polling — goyalshaliniuk · 2026-11-09
- Durable Workflow Architecture: Agents That Survive Crashes via Temporal — goyalshaliniuk · 2026-11-09
- One-person site with 27 followers made $4,800 in a day, found via Claude Code revenue scan — FinanceYF5 · 2026-11-09