Deep Research — CrewAI

Files

• crew_deep_research.py — CrewAI agents + tasks
• requirements-hint.txt — Suggested packages (pin yourself)

Full prompt / config / code

Copy

# crew_deep_research.py — Deep Research (CrewAI)
# Safety: local/OpenAI-compatible endpoint preferred. No keys in source.
# pip install crewai  (pin versions yourself)

from crewai import Agent, Task, Crew, Process

SYSTEM = """You are a careful research desk. Goal: produce a brief, not a verdict.

RULES:
- Use only facts present in the user paste or clearly marked as [UNVERIFIED].
- Separate: Known / Inferred / Unknown / Next checks.
- Cite sources as the user labeled them (URL, doc name, date). Do not invent citations.
- No spending, no shell commands, no credential requests.
- Stop after the brief. Ask one clarifying question only if blocking.

OUTPUT:
1) One-sentence scope
2) Findings (bullets)
3) Open questions
4) Next checks (max 5)
5) Confidence: low|medium|high + why"""

GOAL = "Turn a fuzzy question into a sourced research brief with open questions and next checks \u2014 not a final verdict."

lead = Agent(
    role="Deep Research lead",
    goal=GOAL,
    backstory="Local BotShelf operator. One job, then stop. No spend. No destructive tools.",
    allow_delegation=False,
    verbose=False,
)

reviewer = Agent(
    role="Quality gate",
    goal="Check the lead output against the required sections; list gaps only.",
    backstory="Pedantic reviewer. Does not invent facts.",
    allow_delegation=False,
    verbose=False,
)

produce = Task(
    description=(
        "Follow SYSTEM strictly.\n\nSYSTEM:\n" + SYSTEM +
        "\n\nUSER INPUT:\n{user_input}"
    ),
    expected_output="Complete Deep Research structured result.",
    agent=lead,
)

gate = Task(
    description=(
        "Verify the previous result has all required sections. "
        "Return Pass/Fail + missing sections. Do not rewrite creatively."
    ),
    expected_output="Pass/Fail and gap list.",
    agent=reviewer,
)

crew = Crew(
    agents=[lead, reviewer],
    tasks=[produce, gate],
    process=Process.sequential,
)

def run(user_input: str) -> str:
    return str(crew.kickoff(inputs={"user_input": user_input}))

if __name__ == "__main__":
    print(run("PASTE_EXAMPLE_IN"))

Example in

Question: Why did our signup conversion drop last week?
Sources:
- [A] analytics export 2026-09-01..09-07 (CSV summary pasted)
- [B] changelog: checkout button copy tweak on 09-03
Ask for a brief with next checks.

Example out

Scope: Explain likely drivers of signup conversion drop using only pasted analytics + changelog.
Findings:
- Drop concentrated on mobile checkout step (from A)
- Timing aligns with 09-03 copy tweak (B)
Open questions:
- Was A/B still running?
Next checks:
1) Diff mobile CTA copy pre/post
2) Error rate on payment iframe
Confidence: medium — single-week window.

Model / runtime notes

Two-agent crew (produce + gate). Keep tools disabled unless you add read-only tools on purpose.

Canonical Team

Parent: deep-research · implementation_id: crewai/deep-research

GitHub source

teams/deep-research/crewai · commit eb87e8e049fdf9908e438305ff827c7b617505b5 · license: free-use-at-own-risk

Structural check: PASS · env: structural (structural only — does not set Verified)

Related Library items

• ollama/deep-research
• lm-studio/deep-research
• open-webui/deep-research
• n8n/deep-research

Related marketplace Team pages

Soft thematic links only — not identity merges. Marketplace Teams ≠ Library AI Team records.

• Research Desk (marketplace Team page — separate entity)
• Source Desk (marketplace Team page — separate entity)