eve Chat Template

eve Chat Template is a Next.js chatbot boilerplate from Vercel's eve-examples repository that turns an eve agent into a deployable chat web app, starting with password-only access and scaling to a production multi-user application.

What is the eve Chat Template?

eve Chat Template is a Next.js application scaffold for the eve agent platform (https://eve.dev), published by Vercel in the public eve-examples GitHub repository. It takes an eve agent definition — edited in agent/agent.ts, with behavior in agent/instructions.md and tools in agent/tools/ — and produces a chat web app with sidebar history, Streamdown markdown rendering of assistant text and reasoning, and durable conversation resumption at /chat/[id]. The app deploys through a one-click Deploy with Vercel button and is built with shadcn/ui components, Tailwind CSS, Better Auth, Drizzle, Neon, and Upstash Redis.

Key Features

• Three deployment modes — starter mode uses a shared EVE_CHAT_PASSWORD with a secure session cookie and browser localStorage; production mode uses Sign in with Vercel, Neon persistence, and Upstash Redis rate limiting; local development runs with neither configured. Production takes precedence when its full environment is present, and the app fails closed in a production deployment when no mode is configured.
• Password-first quick start — deploy with only EVE_CHAT_PASSWORD (16+ characters recommended); no database or Marketplace products are needed, and chat history stays in the browser.
• Production upgrade path — ./scripts/setup.sh provisions Neon and Upstash, registers Sign in with Vercel, pulls environment variables, and runs Drizzle migrations for cross-device chat history under lib/db.
• Durable sessions — eve session cursors and event snapshots are saved in either storage mode, letting /chat/[id] resume the same conversation after a refresh.
• Vercel Connect integrations — optional Slack channel route at /eve/v1/slack plus Notion, Linear, and Sentry MCP connections created with vercel connect create; the composer shows its connections menu only when at least one MCP connector is configured.
• Local-first chat features — first-message chat titles are derived locally from the user's prompt, and the sidebar includes delete and new-chat actions.
• Same-origin eve routes — the browser talks to eve through /eve/v1/* routes using useEveAgent() from eve/react, and shadcn/Tailwind components cover messages, tools, human-in-the-loop prompts, and the composer.

Who is it for?

• Independent developers — deploy a working eve chat UI in minutes with a strong password alone, then test an agent's behavior and tools from a browser.
• Product teams shipping multi-user assistants — run ./scripts/setup.sh to provision Neon and Upstash, register Sign in with Vercel, and run migrations, producing an authenticated chat app with server-persisted history and rate limiting.
• Teams wiring eve into collaboration tools — attach Slack, Notion, Linear, or Sentry connectors via vercel connect create and set the resulting UIDs in SLACK_CONNECTOR, NOTION_CONNECTOR, LINEAR_CONNECTOR, and SENTRY_CONNECTOR.
• Developers extending agent code — edit agent/agent.ts and add tools under agent/tools/; custom production domains are supported through the optional BETTER_AUTH_URL variable.

What can you do with the eve Chat Template?

• Prototype an eve agent: click Deploy with Vercel, enter a strong EVE_CHAT_PASSWORD, open the deployed app, and enter that password — chats persist in that browser's localStorage.
• Ship a production chat app: run ./scripts/setup.sh (optionally with --scope team-slug) to provision the database and Redis store, then create tables with pnpm db:migrate (or vercel env run -e production -- pnpm db:migrate).
• Integrate with external systems: create MCP connections for Notion, Linear, and Sentry, or a Slack channel connector, then test them locally because local development falls back to default connector names without editing agent/.
• Resume long conversations: stored session cursors and stream events let a user reload /chat/[id] and continue the same eve conversation, with deletion and new-chat actions in the sidebar.

How does the deployment work?

Quick start takes three steps: click Deploy with Vercel, set a strong EVE_CHAT_PASSWORD, and open the deployed app to enter that password. The production upgrade is driven by ./scripts/setup.sh, which provisions Neon and Upstash, registers Sign in with Vercel, pulls environment variables, and runs migrations. Locally, pnpm install then pnpm dev runs the app without cloud services, and setting EVE_CHAT_PASSWORD in .env.local enforces the same password.

Pros and cons

• Zero-setup starter — deployment requires only the password variable; no database or Marketplace provisioning.
• Gradual scaling — the same repository covers single-operator testing and authenticated multi-user production.
• Starter mode limits — anyone with the password shares the same agent identity and connection grants, and chats stay in one browser, so it is designed for one trusted operator.
• Deliberate exclusions — file uploads, Vercel Blob, guest mode, NextAuth/Auth.js, and AI Elements are intentionally not included.

Alternatives

• Vercel AI Chatbot — another open-source Next.js chat starter from Vercel, built on the AI SDK for direct LLM providers instead of the eve agent runtime.

FAQ

Do I need a database to use this template?

No. Starter mode requires only EVE_CHAT_PASSWORD; chats and eve session cursors live in browser localStorage and are not shared across browsers or users. Neon is required only in production mode, which activates when Neon, Upstash, and all Sign in with Vercel variables are configured.

How do I upgrade from starter to production mode?

Run ./scripts/setup.sh from the repository root, or ./scripts/setup.sh --scope team-slug for a specific team. The script provisions Neon and Upstash, registers Sign in with Vercel, pulls environment variables, and runs migrations; production mode then takes precedence over the password mode.

Can I run the template locally without cloud services?

Yes. Run pnpm install and then pnpm dev. With neither production variables nor EVE_CHAT_PASSWORD set locally, the app uses a local development identity and browser localStorage. Adding EVE_CHAT_PASSWORD (at least 16 characters) to .env.local requires that same password to enter the app.

Which environment variables does production mode need?

Production requires DATABASE_URL, BETTER_AUTH_SECRET, NEXT_PUBLIC_VERCEL_APP_CLIENT_ID, VERCEL_APP_CLIENT_SECRET, UPSTASH_REDIS_REST_URL, UPSTASH_REDIS_REST_TOKEN, KV_REST_API_URL, and KV_REST_API_TOKEN. Optional variables include BETTER_AUTH_URL for custom domains and the Slack, Linear, Notion, and Sentry connector variables.

What does the template intentionally leave out?

File uploads, Vercel Blob, guest mode, NextAuth/Auth.js, and AI Elements are not part of the template. Password-only starter deployments also omit the composer's connections menu because no MCP connectors are configured in that mode.