Long-running execution
Agent runs that keep going past a single request — through a deploy, a timeout, or a client that walks away.
Subako gives your product everything it needs to run AI agents for your users — sessions that keep running even when the connection drops, plus code execution, credentials, and skills. One SDK. Bring your own model.
A model alone stops at an answer. Subako provides the runtime around it — so your agent can keep running, use tools, execute code, and reach the real world.
Agent runs that keep going past a single request — through a deploy, a timeout, or a client that walks away.
Conversation and tool state lives server-side, so a dropped connection resumes where it left off instead of starting over.
Model-written code runs in an isolated guest, with no path back into your own infrastructure.
Store the credentials an agent needs to reach real systems, and broker them at call time rather than handing them to the model.
Package the knowledge and step-by-step procedures an agent should follow into units you can version and reuse.
Bring your own model. Full details at launch.
Subako runs the session store, the retry policy, and the sandbox as one service you connect to in a few lines — so you gain the agent capability without gaining a system to operate.