# Vercel AI Gateway and Vercel Sandbox now available on Hermes Agent

Source: [Vercel](https://vercel.com/changelog/vercel-ai-gateway-and-vercel-sandbox-now-available-on-hermes-agent)  
Feed7 permalink: https://feed7.dev/p/vercel-ai-gateway-and-vercel-sandbox-now-available-on-hermes-agent-04gyfsv  
Published: 2026-08-07T19:00:00.000Z  
Trust: Official Source (official_source)

## Why Included

Hermes can route inference through Vercel’s model gateway and move command execution into an opt-in cloud microVM, separating model access from the machine where the agent runs.

## Source Summary

Hermes now connects to Vercel AI Gateway for **200+ models** with live availability and pricing, no token markup, and centralized request tracking. Its optional Sandbox backend runs commands in an **isolated microVM**.

## Practical Implication

Use the gateway when you want model choice and spend visibility in one place. Enable terminal.backend=vercel_sandbox when agent commands should run away from your workstation; available runtimes include **Node 24, Node 22, and Python 3.13**.

## Agent-Ready Context

Hermes now connects to Vercel AI Gateway for **200+ models** with live availability and pricing, no token markup, and centralized request tracking. Its optional Sandbox backend runs commands in an **isolated microVM**.

Use the gateway when you want model choice and spend visibility in one place. Enable terminal.backend=vercel_sandbox when agent commands should run away from your workstation; available runtimes include **Node 24, Node 22, and Python 3.13**.

Commands remain local by default, so selecting the gateway alone does not isolate execution. Local development with the cloud backend also requires a **VERCEL_OIDC_TOKEN** and a linked Vercel project.

## Connected Context

Feed7 judgment across 395 accumulated Signals:

This combines centralized model access and spend visibility with an optional isolated execution backend for Hermes, but keeps the two controls independent: gateway selection does not move commands off the workstation. It turns sandboxing into an explicit deployment choice with runtime, project-linking, and OIDC prerequisites rather than an automatic property of using a managed model gateway.

- [Qwen 3.8 Max now available on Vercel AI Gateway](https://feed7.dev/p/qwen-3-8-max-now-available-on-vercel-ai-gateway-1ikih0e) — Qwen exemplifies the model routes Hermes can reach through the gateway, while the Hermes integration adds agent-wide tracking and optional execution isolation around such routes.
- [AI Gateway adds unified fast mode support](https://feed7.dev/p/ai-gateway-adds-unified-fast-mode-support-144dq26) — Unified fast mode adds a latency-routing control within the gateway side of Hermes, but remains separate from whether commands execute locally or in Sandbox.
- [chenyme/grok2api](https://feed7.dev/p/grok2api-0jko7ks) — The managed Hermes route contrasts with Grok2API’s unofficial account pooling: it centralizes models and tracking without the candidate’s stated SSO, credential-renewal, and terms risks.
- [Emulated: The Data for Fully Autonomous Software Engineers and Companies — Joseph Wang](https://feed7.dev/p/emulated-the-data-for-fully-autonomous-software-engineers-and-companies-18dpzpf) — Hermes supplies an isolated microVM for command execution, but the evaluation candidate narrows what that proves: a sandbox alone does not cover deployments, incidents, traffic, or distributed failures.

## Context Map

- Layer: infra
- Domains: coding
- Topics: gateways, sandboxing, coding-agents

## Uncertainty

- Commands remain local by default, so selecting the gateway alone does not isolate execution. Local development with the cloud backend also requires a **VERCEL_OIDC_TOKEN** and a linked Vercel project.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.