Skip to main content

​
Agent-as-tool

Convert an agent into a tool that another agent can call: 
// agents/researcher.ts
import { agent } from "veryfront/agent";

export default agent({
  id: "researcher",
  model: "openai/gpt-4o",
  system: "Research topics thoroughly using web search.",
  tools: { webSearch: true },
  maxSteps: 5,
});

// agents/writer.ts
import { agent } from "veryfront/agent";

export default agent({
  id: "writer",
  model: "openai/gpt-4o",
  system: "Write clear, well-structured articles.",
});

// agents/orchestrator.ts
import { agent, getAgentsAsTools } from "veryfront/agent";

export default agent({
  id: "orchestrator",
  model: "openai/gpt-4o",
  system: "You coordinate research and writing. Use the researcher to gather facts, then the writer to produce the article.",
  tools: getAgentsAsTools({
    researcher: "Research a topic using web search",
    writer: "Write an article from research notes",
  }),
  maxSteps: 10,
});
getAgentsAsTools() wraps each agent as a tool. The orchestrator decides when to call each agent based on its system prompt. Each sub-agent runs its own tool loop independently.

​
Single agent-as-tool

For wrapping a single agent: 
import { agentAsTool, getAgent } from "veryfront/agent";

const researcher = getAgent("researcher");
const researchTool = agentAsTool(researcher, "Research a topic using web search");

​
Workflow-based composition

For deterministic multi-agent pipelines, use workflows: 
// workflows/article-pipeline.ts
import { workflow, step, parallel } from "veryfront/workflow";

export default workflow({
  id: "article-pipeline",
  steps: [
    step("research", { agent: "researcher" }),
    parallel("drafts", [
      step("draft-1", { agent: "writer", input: "Style: conversational" }),
      step("draft-2", { agent: "writer", input: "Style: technical" }),
    ]),
    step("select", { agent: "editor" }),
  ],
});

​
When to use which

PatternUse when
Agent-as-toolThe orchestrator decides dynamically which agents to call and in what order
WorkflowThe execution order is known in advance (sequential, parallel, or branching)
Agent-as-tool is more flexible but harder to predict. Workflows are deterministic and easier to debug.

​
Agent registry

All agents in agents/ are registered automatically. Access them programmatically: 
import { getAgent, getAllAgentIds } from "veryfront/agent";

const ids = getAllAgentIds();    // ["assistant", "researcher", "writer"]
const agent = getAgent("writer"); // Get a specific agent

​
Next

  • Providers: configure OpenAI, Anthropic, and Google
  • Middleware: add auth and rate limiting to your agents