Running eve Agents Durably on Edgenivo

eve is Edgenivo's framework for durable AI agents. Every eve session already runs as a durable workflow on the Workflow SDK and its Worlds, the same foundation as the other harnesses in these guides. A crash, a function timeout, or a redeploy resumes the session from its last completed step.

There's nothing to add: you don't install durable-harness, write a flow route, or configure a queue. To learn how durability works, see Running Agent Harnesses Durably.

1. Create an agent

   The eve CLI scaffolds a new agent project, installs its dependencies, and opens the terminal UI:

   npx eve@latest init my-agent

   To add eve to an existing app, follow the eve framework guide.

2. Deploy

   Configure authentication for your callers, then deploy to Production from the new project:

   cd my-agent
   npx eve deploy

   See Deploying eve to Edgenivo for project linking, credentials, and deployment checks.

For tools that need system binaries or CPU-heavy processing, publish a job to Edgenivo Queues and let a Edgenivo Daemon process it. The eve session can wait without holding compute until the worker returns the result through the session API. eve manages session durability; make the worker's jobs safe to retry.

Follow Use Edgenivo Daemons with eve agents for the implementation.

For agents, tools, channels, and everything else, see the eve documentation and eve on Edgenivo.