Agents
Agents compose models, knowledge and Memory under your policies. The agent runtime is coming soon (roadmap V2), and multi-agent orchestration will compose several governed agents under one policy.
Agents, MCP & Tools
Build agents that stay on leash
When it arrives, the agent runtime composes models, knowledge, Memory and MCP tools into governed agents. Your own enterprise systems will plug in as tools with the same identity, guardrails and routing as native capabilities.
Three capabilities, one policy
Everything an agent does passes the controls that govern every other platform request.
MCP built in
MCP is built into the design: your tools and data sources will connect as MCP servers, while each tool call runs under the same identity, guardrails and routing as a native model call.
Guardrails on the whole path
Prompt and injection detection plus policy checks will run before and after every model and tool action, with the decisions logged for review.
Start building agents
Bring your models, your knowledge and your tools; the platform keeps the whole path governed and auditable.
Related platform capabilities
Knowledge, Brain & Memory
Bring your organization's knowledge into every inference and agent.
Coming soon
AI Router
One endpoint that routes every request to the model and region you choose.
Available now
Observability & FinOps
Usage, cost and quality per model, region and team, at a glance.
Coming soon