APIs — devopsmantra wiki

347 saved links, 2017–2026 (peak: 2023). This tag’s biography mirrors your own: it opens with REST endpoint design, Azure API Management, and serverless APIs from the DevOps years — then in 2023 the word “API” quietly changes meaning to model endpoint, and developers.openai.com becomes a top source. The newest saves complete the arc: you’re no longer just designing APIs for humans, you’re designing them for agents — MCP servers, function-calling schemas, and gateways that govern what a model is allowed to touch.

Related: System Design & Architecture · AI Agents · LLMs · Azure · Security

API design — the classics#

The pre-LLM canon. Everything here still applies; the consumers just got stranger.

How to (and how not to) design REST APIs — your note: “overall a very reasonable set of guidelines,” even if you weren’t buying strings-for-all-IDs.
How Slack designs its APIs — saved with the principle “I listen to API design guidance from those with good ecosystems.”
Zalando’s RESTful API guidelines — the enterprise-grade rulebook, still worth saving in 2025.
Kelsey Hightower on producing elite API design (video) — “loads of advice for design in general.”
REST API best practices and endpoint design · The Art of Writing Amazing REST APIs · the REST design cheat sheet.
A complete guide to the API-first approach and API architectural styles — the strategy layer.
Mastering API Architecture (O’Reilly) — saved the day the authors finished writing it.

Beyond REST — GraphQL, gRPC, webhooks#

How Netflix migrated to GraphQL safely — the migration case study.
Uber’s chat architecture on GraphQL subscriptions — reliability and scale as the drivers.
the gRPC introduction — “gRPC has one of the coolest documentation on their website.”
State of Webhooks 2023 — 83% of API providers ship webhooks; most skip the best practices.
paste an entire GraphQL schema into Claude and ask for queries — Simon Willison, 2024; the moment the two halves of this tag met.

The LLM API era#

Where the tag pivoted. The Responses API migration is your most-tracked storyline.

Simon Willison: Responses API vs Chat Completions — the launch-day analysis, and the Latent Space interview on OpenAI’s Agents Platform — the whole API suite refreshed “for the year of Agents.”
new tools in the Responses API — o3/o4-mini calling tools inside their chain-of-thought; PDF file inputs landed the same season.
the Chat Completions → Responses migration kit and the ported azure-search-openai-demo — when the samples repos migrated, you saved the receipts.
OpenAI’s function-calling guide · function calling crash course (video) · function calling with the Gemini API.
how streaming LLM APIs actually work — Willison dissects OpenAI, Anthropic, and Gemini streaming; the Batch API’s 50%-off deal for anything that can wait 24 hours.
Anthropic: API fundamentals course · the Claude API development guide · notes on the Citations API — interesting precisely because you build RAG.
Gemini: the Gemini API cookbook · function calling plus built-in tools in a single call (“This simplifies a few of my own agents”) · Gemini Skills.
The honest counterweight: FineTuneBench — commercial fine-tuning APIs mostly don’t teach models new facts.

MCP & agent-facing APIs#

The current frontier: APIs whose primary consumer is a model. This is the layer your architect thinking lives in.

MCP vs APIs: when to use which for agent development — saved with your own field report: “I’m using both techniques for an agent I’ve built, and see reasons to use APIs directly.”
MCP vs APIs: the new standard for AI integration — the framing piece; OpenMCP — convert web APIs into MCP servers for the conversion path.
building MCP servers for ChatGPT and API integrations · the Responses API MCP tool guide.
Vercel’s agent-friendly APIs course — build an API, document it for agents; “finally, someone is reading the docs.”
giving ADK an OpenAPI tool — “giving an AI an API as a ‘tool’ is a fairly powerful pattern. Heck, that’s basically what MCP is.”
Google’s Developer Knowledge API + MCP server — a programmatic source of truth so LLMs stop citing stale docs.
Agents, APIs, and the Next Layer of the Internet — the big-picture essay.
9 tips for reducing API latency in agentic systems — ops discipline applied to agents.
Two grounding shots: multi-agent AI is the new microservices — “not everything should be an agent. Write a script. Keep that API.” — and a coding agent is a while loop, an API call, and a few Python functions.

Gateways, governance & security#

The classic APIM thread evolving into AI-gateway architecture — arguably this tag’s most career-relevant seam.

an enterprise AI gateway with Azure API Management + AI Foundry and a Foundry agent harness: tool governance on Azure API Gateway — the 2026 pattern: the gateway as the agent’s policy layer.
the busy platform engineer’s guide to API gateways — treat the gateway as a product; choosing an API gateway for microservices.
load balancer vs reverse proxy vs API gateway — saved with a dissent: “(I don’t agree with the definition of each).”
APIOps + IaC reshaped an API strategy — taking developers out of the portal.
awesome-api-security — the tools-and-resources list.
how Google AI Studio proxies Gemini API requests — so your key doesn’t ship in client-side code; the API-key hygiene lesson.
The Azure APIM heritage: an APIM sinkhole with App Gateway + Terraform · APIM with developer portal via Azure DevOps CI/CD · MuleSoft Flex Gateway + LLMs for governed API intelligence.

Tooling & the workbench#