OptionalagentUrl?: stringURL the executor uses to reach this agent's data layer.
Auto-derived when the agent runs on its own server (mountOnStandaloneServer).
Required when the agent is mounted on an external framework (Express/Fastify/NestJS),
since the agent cannot know the host application's address.
Optionalai?: Bring your own AI provider instead of Forest's AI server. Omit ai entirely to keep using
Forest's server. bedrock takes no apiKey: credentials come from the AWS credential chain
(IAM role, AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY, shared profile). region falls back to
AWS_REGION / AWS_DEFAULT_REGION, but never to the profile's own region.
Optionaldatabase?: { uri?: string; [option: string]: unknown }Database connection used to persist workflow run state. Accepts a connection URI or a
Sequelize options object. The agent throws at startup if it is omitted (unless inMemory is
set).
OptionalencryptionKey?: stringHKDF secret encrypting OAuth-protected MCP connector credentials at rest. Only needed if you
use OAuth-protected MCP connectors; without it their credential deposits return 503. Embed-time
equivalent of the standalone FOREST_EXECUTOR_ENCRYPTION_KEY env var.
OptionalinMemory?: booleanUse an in-memory run store instead of a database. No database is required, but runs are lost
when the process restarts, so it is not meant for production. Mutually exclusive with
database.
Optionalport?: numberLoopback port the embedded executor listens on; the agent proxies to it internally.
Defaults to 3400.
Options for an embedded workflow executor, started in the same process as the agent through
agent.addWorkflowExecutor().