Multi-Agent

multi_agent answers one question with a team of web research agents. Each agent researches a different angle, then their findings are synthesized into one structured result.

Use multi_agent for a single question that deserves depth or breadth. To run one agent per row of a table, use agent_map.

Examples

A single question

Pass an empty DataFrame for a standalone question.

import pandas as pd
from futuresearch.ops import multi_agent

result = await multi_agent(
    task="Research the current state of formal verification for AI systems",
    input=pd.DataFrame(),
)
print(result.data.head())

The team explores different angles in parallel, then synthesizes one answer. By default the result has a single answer column.

Structured output

Pass a response_schema (a JSON Schema object) to get a multi-field result.

import pandas as pd
from futuresearch.ops import multi_agent

result = await multi_agent(
    task="Research the current state of formal verification for AI systems",
    input=pd.DataFrame(),
    response_schema={
        "type": "object",
        "properties": {
            "overall_maturity": {"type": "string"},
            "what_works": {"type": "string"},
            "key_limitations": {"type": "string"},
            "trajectory": {"type": "string"},
        },
        "required": ["overall_maturity", "what_works", "key_limitations", "trajectory"],
    },
)

Explicit directions

By default the research angles are generated for you. Pass directions to set them yourself, up to 6. Each direction is a full brief, not a short title.

result = await multi_agent(
    task="Research the major prediction-market platforms",
    input=pd.DataFrame(),
    directions=[
        "Research Polymarket: mechanism, volume, regulatory status, user base.",
        "Research Kalshi: mechanism, volume, regulatory status, user base.",
        "Research Metaculus and other forecasting-aggregation platforms.",
    ],
)

Generate a list

Set return_list=True to turn one question into a list. The response_schema describes a single item; the result has one row per item, with an _expand_index column.

import pandas as pd
from futuresearch.ops import multi_agent

result = await multi_agent(
    task="Find startups that sell training data, benchmarks, or RL environments to frontier AI labs",
    input=pd.DataFrame(),
    return_list=True,
    response_schema={
        "type": "object",
        "properties": {
            "company_name": {"type": "string"},
            "category": {"type": "string"},
            "funding_stage": {"type": "string"},
        },
        "required": ["company_name", "category", "funding_stage"],
    },
)
print(result.data)  # one row per company

Parameters

Effort levels

The effort level sets the number of research agents on the question.

• LOW: 3 agents
• MEDIUM: 4 agents (default)
• HIGH: 6 agents

Cost and runtime

Cost scales with effort. Across production runs, a single question costs roughly:

Via MCP

MCP tool: futuresearch_multi_agent

Related docs

Guides

• Research a Question with a Team of Agents

Case Studies

• Find Startups Selling to AI Labs
• Research Formal Verification for AI

Reference

• agent_map to run one agent per row of a table