Agentic AI
4 articles on Agentic AI.
Designing Async, Long-Running APIs for AI Agents
AI agents kick off tasks that run for minutes — synchronous request/response breaks down fast. A practical guide to the async job pattern: 202 + status URLs, polling vs webhooks vs streaming, durable queues, and idempotent resumption.
May 30, 2026·6 min readMCP vs Direct API Calls: The Token-Efficiency Debate (2026)
MCP exploded to 10,000+ servers, but in 2026 many teams are moving back to direct API calls and CLIs over token cost — ~200 tokens per CLI command vs 32,000–82,000 for MCP. A clear-eyed breakdown of when MCP is worth it and when it isn't.
May 26, 2026·5 min readBuilding Backends for AI Agents: Idempotency, Retries & State (2026)
AI agents retry, run for minutes, and call your APIs in unpredictable loops. The backend is where agent reliability lives. A practical guide to idempotency, safe retries, durable state, and observability for agent-facing systems in 2026.
May 22, 2026·7 min readBuilding AI-Native Backends: Architecture for Autonomous Agents in 2026
Complete guide to designing backend systems for AI agents - event-driven architectures, MCP protocol, vector databases, agent governance, and production patterns for 2026.
December 19, 2025·11 min read