Skip to main content

Documentation Index

Fetch the complete documentation index at: https://veryfront.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

​
Create an agent

Define an AI agent in a fresh Veryfront project, send it a message, and read back the response. This guide is the smallest end-to-end agent you can build with Veryfront; deeper topics — tools, memory, skills, hosting, streaming — live in Agents and the AI guides that follow it.

​
Prerequisites

  • Veryfront installed and a project created with Quickstart. The ai-agent template gives you the file layout below by default.
  • An agents/ directory in the project root. If you started from the minimal template, create one: mkdir agents.
  • A provider configured for inference. The simplest path is to set OPENAI_API_KEY or ANTHROPIC_API_KEY in .env; the framework picks the matching provider automatically. See Providers for other options.

​
Define the agent

Create agents/assistant.ts: 
// agents/assistant.ts
import { agent } from "veryfront/agent";

export default agent({
  id: "assistant",
  system: "You are a concise assistant. Answer in one short paragraph.",
});
The file name (without extension) becomes the agent id. Veryfront auto-discovers every file under agents/ at startup, so there is nothing else to register. For a one-shot persona without TypeScript, you can write the same agent as agents/assistant.md: 
---
name: Assistant
description: Concise general-purpose assistant
---

You are a concise assistant. Answer in one short paragraph.
Both forms register the same assistant id and are interchangeable from the call sites below.

​
Invoke the agent

From any server-side context (an API route, getServerData, a workflow step, a CLI command), resolve the agent by id and call generate: 
// app/api/ask/route.ts
import { getAgent } from "veryfront/agent";

export async function POST(request: Request) {
  const { question } = await request.json();
  const assistant = getAgent("assistant");

  const result = await assistant.generate({ input: question });

  return Response.json({
    answer: result.text,
    toolCalls: result.toolCalls,
    usage: result.usage,
  });
}
App-router handlers receive the raw Request directly. If you prefer the pages router, the same route lives at pages/api/ask.ts and receives an APIContext (ctx.request.json(), ctx.json(...)); see API routes. generate returns the full response. Use stream when you want to send chunks back over Server-Sent Events; the Memory and streaming guide covers that path.

​
Run it

Start the dev server: 
veryfront dev
Send a request from another terminal: 
curl -X POST http://localhost:3000/api/ask \
  -H "content-type: application/json" \
  -d '{"question":"What is Veryfront in one sentence?"}'

​
Verify it worked

You should see a JSON response whose answer is a short paragraph from the model. The usage object reports input and output tokens. If the dev server logs an error mentioning a missing provider, recheck that OPENAI_API_KEY (or your provider’s variable) is exported in the shell that runs veryfront dev.

​
Next

  • Agents: full agent surface — tools, dynamic system prompts, multi-step runs, memory, AG-UI handlers, hosted runs.
  • Tools: let the agent call typed functions.
  • Chat UI: drop a streaming chat interface into a React page that talks to this agent.