Agent Stream

Govern every model call your agents make. Route across providers, enforce guardrails inline, and keep every execution attributable.

When agents move into production, model execution spreads across multiple providers, tools, and data sources, each with its own limits, cost profile, and failure modes. Teams end up rebuilding routing, guardrails, and cost controls separately for every integration.

Agent Stream is an execution layer that sits between your agents and the model providers they call, applying routing, guardrails, and cost policy to every execution from one configuration layer. You point the agent at the Agent Stream URL, and it governs execution from there. You change only the caller’s target URL, and behaviour stays the same.

Every execution becomes governed, cost-accountable, and auditable. Get started →

Overview→

What Agent Stream is, how it governs model execution across providers with routing, guardrails, budgets, and observability, and how to adopt it progressively.

Quickstart→

Run Agent Stream, route your first execution through it, and add guardrails from the dashboard.

Core Concepts→

Concept pages covering Agent Stream’s architecture, request pipeline, guardrails, cost governance, attribution, and how it extends governed model access to editors and enterprise document stores.

Guides→

Step-by-step guides for real configuration and operational tasks in Agent Stream.

Reference→

Field-level reference for Agent Stream surfaces, routing policies, provider configuration, guardrails, settings, and operations.

Frequently asked questions→

Common questions about routing, providers, guardrails, cost accounting, configuration, and deployment.