AI Platform Docs
Build AI systems like software. Define prompts, connect tools, compose agents, run them through a managed runtime, and inspect what actually happened in production.
What this documentation covers
These docs are organized around the objects that make up the platform: prompts, tools, agents, voice agents, runs, realtime sessions, service keys, and API keys. They also document the rest of the app where it matters, including onboarding, billing, users and groups, profile settings, and support.
Guides
API conventions
Learn how organization scoping, request shapes, timestamps, and list responses work.
Client-side API calls
Future-facing guidance for browser-safe access patterns and delegated execution.
Resources
Prompts
Define reusable AI behavior with versions, variables, layers, models, and attached tools.
Tools
Connect APIs, databases, and internal services so models can call external capabilities safely.
Agents
Compose prompts and tools into multi-step workflows with transitions, branching, and inspectable runs.
Voice agents
Wrap prompt versions for realtime conversations with voice configuration and agent handoffs.
Runs
Inspect prompt and agent execution records, including tokens, latency, cost, composition, and tool calls.
Realtime sessions
Review live session timelines, handoffs, tool calls, and graph snapshots for voice workflows.
Service keys
Backend automation credentials for org-scoped AI platform management and runtime inspection.
API keys
Developer-managed application credentials for the intended public API integration surface.
SDK roadmap
Node.js
The planned first-party TypeScript SDK for backend services, workers, and CI automation.
Ruby
Planned SDK guidance for Rails apps integrating prompts, runs, and service-key automation.
PHP
Planned SDK coverage for backend integrations that need prompt execution and observability APIs.
