> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentium.in/llms.txt
> Use this file to discover all available pages before exploring further.

# Collaborate Mode

> Iterative rounds between agents. maxRounds config. Example with iterative refinement.

# Collaborate Mode

In **Collaborate** mode, agents work in **iterative rounds**. Each agent can build on the previous agent's output. The process continues until a stopping condition or `maxRounds` is reached.

***

## How It Works

<Steps>
  <Step title="Round 1">
    Agent A processes the initial input and produces output.
  </Step>

  <Step title="Round 2">
    Agent B receives A's output, adds its contribution.
  </Step>

  <Step title="Subsequent rounds">
    Agents continue passing work until done or maxRounds.
  </Step>

  <Step title="Final output">
    The team returns the last (or merged) result.
  </Step>
</Steps>

***

## maxRounds Config

<ParamField path="maxRounds" type="number" required={false}>
  Maximum number of collaboration rounds. Prevents infinite loops. Set based on your workflow (e.g., 3–5 for typical pipelines).
</ParamField>

***

## Example: Iterative Refinement

<CodeGroup>
  <CodeGroup.Item title="collaborate-team.ts">
    ```typescript theme={null}
    import { Team, TeamMode, Agent, openai } from "@agentium/core";

    const drafter = new Agent({
      name: "Drafter",
      model: openai("gpt-4o-mini"),
      instructions: "Create initial drafts. Be creative and comprehensive.",
    });

    const editor = new Agent({
      name: "Editor",
      model: openai("gpt-4o-mini"),
      instructions: "Improve drafts: fix clarity, structure, and flow. Keep the core ideas.",
    });

    const polisher = new Agent({
      name: "Polisher",
      model: openai("gpt-4o-mini"),
      instructions: "Final pass: tighten language, fix grammar, ensure consistency.",
    });

    const writingTeam = new Team({
      name: "Writing Pipeline",
      mode: TeamMode.Collaborate,
      model: openai("gpt-4o"),
      members: [drafter, editor, polisher],
      maxRounds: 3,
      instructions: `
        Agents collaborate in sequence:
        1. Drafter creates the initial version.
        2. Editor refines structure and clarity.
        3. Polisher does final polish.
        
        Each agent builds on the previous output. Stop when the result is publication-ready.
      `,
    });

    const output = await writingTeam.run(
      "Write a blog post about the benefits of TypeScript for large codebases."
    );
    console.log(output.text);
    ```
  </CodeGroup.Item>
</CodeGroup>

***

## When to Use Collaborate

<CardGroup cols={2}>
  <Card title="Iterative refinement" icon="rotate">
    Draft → Edit → Polish pipelines.
  </Card>

  <Card title="Peer review" icon="user-check">
    Agent A proposes, Agent B reviews and suggests changes.
  </Card>

  <Card title="Multi-stage generation" icon="layer-group">
    Each agent adds a layer of value.
  </Card>

  <Card title="Convergence" icon="bullseye">
    Agents iterate until output meets criteria.
  </Card>
</CardGroup>

***

## Tips

<Accordion title="Setting maxRounds">
  Start with 3–5 rounds. Too few may cut off useful iteration; too many increases cost and latency.
</Accordion>

<Accordion title="Round order">
  The order of members matters. Place agents in the sequence that matches your workflow (e.g., drafter first, polisher last).
</Accordion>