Agent runtime protocol

A Stackbone container runs a workspace: one or more in-process agents (multi-turn chat, selected by name) plus durable workflows. You do not hand-roll an HTTP server. The Stackbone workspace runtime owns the HTTP surface; you author files (deep-agents/<name>/index.ts for each agent, workflows/<name>.workflow.ts for each workflow) with @stackbone/sdk. The runtime serves chat, the model catalog and the workflow routes for you.

This page is the behavioural contract: the HTTP shapes your workspace serves, the credentials they expect, and the env the runtime injects. For the underlying model, see Agents, Workflows, and Stackbone Connect.

What a workspace serves

A workspace exposes four layers of HTTP, all on the same server:

Layer Routes Purpose
Agent chat POST /openai/v1/chat/completions, GET /openai/v1/models, POST /anthropic/v1/messages, GET /anthropic/v1/models talk to any agent as an OpenAI or Anthropic client
Agent chat (AG-UI) POST /agui/v1/agents/:name drive an agent from an AG-UI client
Workflow runtime POST /api/workflows/:name/start, POST /api/workflows/:name/chat, GET /api/workflows, GET /api/workflows/:name/schema, GET /api/discovery start and inspect durable workflow runs
Lifecycle GET /live, GET /health, GET /api/contract what an orchestrator probes, plus the capability handshake

You consume all four as a stable contract and implement none of them. The framework produces every route from the files you author.

BUILT WITH ❤️ FROM CANADA AND SPAIN