Vercel AI Gateway and Vercel Sandbox now available on Hermes Agent
Vercel has integrated its AI Gateway and Vercel Sandbox with Hermes Agent, enabling cloud-based model inference and isolated command execution in microVMs.
Hermes Agent now supports Vercel AI Gateway as its inference layer, allowing users to access over 200 models without additional token markup. Requests are logged in the AI Gateway dashboard alongside other usage data, and the model picker dynamically displays live availability and pricing. The integration is designed to streamline model selection and cost tracking within a unified interface. Hermes Agent documentation provides a step-by-step guide for configuration and usage.
Vercel Sandbox operates as an opt-in backend for Hermes Agent, enabling commands to run in an isolated cloud microVM instead of locally. When enabled, the workspace root is set to /vercel/sandbox, and terminal.backend must be configured to vercel_sandbox. This setup ensures commands execute in a secure, reproducible environment with supported runtimes including node24 (default), node22, and python3.13.
For local development, users must authenticate using VERCEL_OIDC_TOKEN, obtained via vercel link and vercel env pull commands. This token facilitates secure access to Vercel’s services during the development phase. The process ensures developers can test and iterate on agent commands before deploying to the cloud microVM environment.
To begin using the integrated features, developers can consult the Hermes documentation for agent-specific guidance and the Vercel Sandbox documentation for microVM configuration. Alternatively, users can explore available AI Gateway models directly through the provided resources. The integration aims to enhance flexibility and scalability for agent-based workflows.