executor

import { connectAgents } from "@nylorun/agents/executor";

Guides: Executors, App server.

connectAgentsCopy Anchor Link

connectAgents(options: ConnectOptions): AgentConnection

Saves every definition, registers an executor for each runnable, and opens an authenticated event stream. When the Runtime offers an action, the executor claims it, runs your implementation, and returns the result. It runs agent tools and hooks, and workflow tool, fn, and verify nodes. Sandbox tools run in the Runtime instead.

Returns an AgentConnection:

Throws (via ready) ConnectionError when no connection is found, and IncompatibleRuntimeError on version skew.

const connection = connectAgents({ agents, onError: console.error });
await connection.ready;

CredentialsCopy Anchor Link

With an application key, PUT /v1/executors registers a derived token for each (application key, Tenant, runnable id). Tokens are never stored. Claims send only { requestId, implementationVersion }. With NYLORUN_EXECUTOR_KEY, the executor uses that pre-provisioned, single-runnable credential.

Delegation limitsCopy Anchor Link

Agents used as tools cannot declare approval. ctx.ask, ctx.approve, ctx.sleep, and ctx.waitFor inside a delegated agent fail with delegation.interaction-unsupported.

deriveExecutorTokenCopy Anchor Link

Computes the same derived executor token the SDK registers. Most apps never call it directly.