Vercel eve at a glance
eve 1.0, Vercel's open-source agent framework, became generally available on November 17, 2026, after a June public preview and 230+ releases: you own the harness, agents execute durably on the Workflow SDK, and Vercel's hosted median turn latency fell from 3.036s to 414ms.
Ask your AI about this story
Ask what changed in eve and what to check before using it in your business. Weigh pricing, the integrations it needs, setup effort, reliability, human oversight and how locked in I would become.
What changed with Vercel eve
Vercel made eve 1.0 generally available on November 17, 2026. eve is Vercel's open-source agent framework for building production-ready agents: its harness — the loop that calls models, runs tools and delegates work to subagents — is owned by the developer, who chooses the model, swaps tools and sets which actions need approval. The framework runs on the open-source Workflow SDK for durable execution, saving progress as the agent works so a run resumes from its last completed step after a crash. For managed hosting, agents deploy to Vercel Workflows and Vercel Sandbox, and an agent and its site can ship as one Vercel project, with native integrations for Next.js, Nuxt, SvelteKit, TanStack Start and Vite.
Sessions are multiplayer, and eve connects natively to Slack, Microsoft Teams, Discord, Telegram, GitHub, Linear, Twilio, Web Chat and iMessage; the Chat SDK adds WhatsApp, Messenger, Google Chat and email, and a Slack ambient mode lets an agent listen in chosen channels and speak only when useful. One deployment can serve many tenants with per-tenant credentials, memory and approval policies, agents keep memory across sessions, and OpenTelemetry tracing plus the Agent Runs dashboard show every run. The framework is open source; managed runs bill through Vercel Workflows and Sandbox usage. By Vercel's figures, hosted median turn latency fell from 3.036s in the September beta to 414ms (633ms p95) at 1.0, and a 50-turn session's estimated Workflows cost fell from 5.4¢ to 1.3¢, about 76% less. Rillet and Clay already run eve in production.
Vercel eve in numbers
Values are copied from the linked source on the date shown. They are not IA Compa measurements.
- Median turn latency, hosted (Vercel's figures)as of · Source for Median turn latency, hosted (Vercel's figures): vercel.com (opens in a new tab)
- 414 ms
Where to compare Vercel eve
Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.
Why Vercel eve matters for businesses
Why it matters: if your business builds agents for clients or for its own operations, eve gives you the production plumbing — durable runs, human approvals, multi-tenant isolation and presence in the chat apps your customers already use — without handing control of the agent's behaviour to a platform. Because the framework is open source and the harness is yours, you can self-host, switch models and audit exactly what an agent may do; and with per-turn costs falling as sessions get longer, leaving an agent running on a long task gets cheaper to justify.
Put it to work: AI Agents
Get AI News by email
Verified AI stories, each with its official source, on Monday and Thursday mornings. Add breaking news to hear about the biggest ones as soon as we publish them.
Every story links its official source
Free, in English or Spanish
Unsubscribe with one click, any time
Sources for this story
- Introducing eve 1.0 (opens in a new tab)Official
Vercel · vercel.com
- eve — the agent framework (opens in a new tab)Official
Vercel · eve.dev
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #eve
- #vercel
- #agent-framework
- #ai-agents
- #open-source
- #workflow-sdk
- #durable-agents


