Agent-First APIs

Agent-first APIs are model-provider interfaces designed for stateful, tool-heavy, long-running agent work rather than single request-response model calls.

Key points

  • Google's Interactions API reached general availability and became Google's primary interface for Gemini models and agents [src-143].
  • The API uses typed steps and supports stateful agent workflows, managed agents, background execution, and past-interaction retrieval [src-143].
  • This pattern matters because agent reliability depends on platform primitives such as state, traces, resumability, tool calls, and execution boundaries, not only model quality.

Related entities

Related concepts

Source references

  • [src-143] Ali Cevik / Google – "Interactions API: our primary interface for Gemini models and agents" (2026-06-22)

Robin Cartier perspective

This page is part of Robin Cartier's working AI knowledge graph: a practical research layer for production AI, recommendation systems, experimentation, GEO, and agentic web readiness.

The useful next step is to connect this concept back to applied product leadership and operating models.

Recommended next

Keep reading from this thread

From 500 indexed pages and articles.

  1. Wiki concept Agent Tool Latency Bottleneck The pattern where a model becomes fast enough that the slow part of the workflow is no Related by apis
  2. Wiki concept Long-Running Agents Agent workflows designed to keep working across long horizons, pauses, approvals, external events, and restarts without losing state or confusing old Related by google
  3. Wiki concept Multi Model Agent Platforms Enterprise agent control planes that make multiple first-party and partner models available behind common governance, quota, and orchestration layers. Readers have engaged with this next