M&A Due Diligence Swarm

pip install requests python-dotenv

SWARMS_API_KEY=your_api_key_here

import json
import os

import requests
from dotenv import load_dotenv

load_dotenv()

API_KEY = os.getenv("SWARMS_API_KEY")
BASE_URL = "https://api.swarms.world"

headers = {"x-api-key": API_KEY, "Content-Type": "application/json"}

target_brief = """
Target: NorthArc Logistics, Inc.
Sector: Mid-market last-mile B2B logistics, U.S. Midwest
Stage: Profitable, founder-owned, 12 years old

Headline financials (target-provided, unaudited):
- TTM revenue: $84M, growing 11% YoY
- TTM EBITDA: $9.8M (11.7% margin)
- Net working capital: $6.1M
- Long-term debt: $14M (senior secured, 7.25% fixed)
- Customer concentration: top 3 customers = 47% of revenue

Asking price: $78M cash-free debt-free, plus working-capital peg.

Strategic context:
- Acquirer is a PE-backed national logistics roll-up at $400M revenue.
- Two prior LOIs on the same target broke in diligence (reason unknown).
- Founder will roll 20% equity and stay for a 2-year earnout.

Known issues raised by sell-side banker:
- Two trucks involved in 2023 multi-vehicle accident, litigation ongoing.
- Recent state sales-tax audit closed, but federal nexus question
  raised by acquirer's tax advisor.
- Custom dispatch software built in-house; no vendor, two engineers
  on payroll maintain it.
- One union shop (one of three terminals), CBA expires in 14 months.

Deliverable: structured deal memo with go / no-go / conditional
recommendation and the three deal-breaker items to clear before LOI.
"""

payload = {
    "name": "M&A Due Diligence Swarm",
    "description": (
        "Cross-functional diligence: Deal Lead synthesizes Financial, "
        "Legal, Tax, and Technical findings into a single deal memo."
    ),
    "swarm_type": "HierarchicalSwarm",
    "max_loops": 1,
    "task": (
        "Conduct a first-pass M&A diligence on the target below. Each "
        "specialist should produce findings strictly within their domain. "
        "The Deal Lead must produce a structured deal memo with: "
        "(1) top red flags ranked by severity, (2) financial highlights "
        "and quality-of-earnings concerns, (3) legal, tax, and technical "
        "risks, (4) a go / no-go / conditional recommendation with the "
        "three specific items that must be cleared before signing an LOI, "
        "and (5) suggested valuation adjustments.\n\n"
        f"TARGET BRIEF:\n{target_brief}"
    ),
    "agents": [
        {
            "agent_name": "Deal Lead",
            "description": (
                "Diligence partner. Reviews each workstream, reconciles "
                "conflicts, and writes the single deal memo."
            ),
            "system_prompt": (
                "You are the Deal Lead on this M&A engagement. You do NOT "
                "redo the specialists' work. Your job is to: "
                "(1) rank red flags across all four workstreams by deal "
                "impact, (2) reconcile disagreements, (3) deliver a "
                "go / no-go / conditional recommendation, (4) name the "
                "three specific items that must clear before signing an "
                "LOI, and (5) propose any valuation adjustments. Output "
                "a structured memo. Be decisive. No hedging without "
                "naming the specific data you would need to decide."
            ),
            "model_name": "gpt-4.1",
            "role": "coordinator",
            "max_loops": 1,
            "max_tokens": 8192,
            "temperature": 0.3,
        },
        {
            "agent_name": "Financial Analyst",
            "description": "Quality of earnings, working capital, debt.",
            "system_prompt": (
                "You are an M&A financial analyst. Focus on quality of "
                "earnings, working-capital normalizations, customer "
                "concentration risk, EBITDA add-backs that are likely "
                "to be challenged, and debt structure. Identify the top "
                "QoE concerns and the specific schedules you would need "
                "from the sell-side data room to clear them. Do not "
                "comment on legal, tax, or technology matters."
            ),
            "model_name": "gpt-4.1",
            "role": "worker",
            "max_loops": 1,
            "max_tokens": 4096,
            "temperature": 0.3,
        },
        {
            "agent_name": "Legal Counsel",
            "description": "Corporate, litigation, employment, regulatory.",
            "system_prompt": (
                "You are M&A legal counsel. Focus on litigation exposure, "
                "the union CBA expiring in 14 months, employment and "
                "change-of-control issues, material contracts, customer "
                "assignability, and regulatory matters. Identify the top "
                "legal red flags and the documents you would request "
                "first. Do not comment on financial, tax, or technical "
                "matters."
            ),
            "model_name": "gpt-4.1",
            "role": "worker",
            "max_loops": 1,
            "max_tokens": 4096,
            "temperature": 0.3,
        },
        {
            "agent_name": "Tax Counsel",
            "description": "Federal, state, sales/use, structuring.",
            "system_prompt": (
                "You are M&A tax counsel. Focus on the federal sales-tax "
                "nexus question, multi-state exposure for a logistics "
                "operator, deal-structure implications (stock vs asset, "
                "338(h)(10), F-reorg), the founder's rollover equity, "
                "and any indemnity / escrow recommendations for tax "
                "contingencies. Quantify exposure where possible. Do not "
                "comment on financial, legal, or technical matters."
            ),
            "model_name": "gpt-4.1",
            "role": "worker",
            "max_loops": 1,
            "max_tokens": 4096,
            "temperature": 0.3,
        },
        {
            "agent_name": "Technical Auditor",
            "description": "Tech stack, key-person risk, integration.",
            "system_prompt": (
                "You are a technical diligence auditor. Focus on the "
                "in-house dispatch software: key-person risk on the two "
                "engineers, code ownership and IP assignment, scalability "
                "to the acquirer's $400M roll-up, integration cost, and "
                "the build-vs-buy decision post-close. Identify the top "
                "technical risks and the artifacts you would request "
                "(architecture docs, code review, runbooks). Do not "
                "comment on financial, legal, or tax matters."
            ),
            "model_name": "gpt-4.1",
            "role": "worker",
            "max_loops": 1,
            "max_tokens": 4096,
            "temperature": 0.3,
        },
    ],
}

response = requests.post(
    f"{BASE_URL}/v1/swarm/completions",
    headers=headers,
    json=payload,
    timeout=600,
)

result = response.json()

for output in result.get("output", []):
    print("=" * 60)
    print(output["role"])
    print("=" * 60)
    content = output["content"]
    if isinstance(content, list):
        content = " ".join(str(c) for c in content)
    print(str(content)[:600] + "...")

print(f"\nTotal cost: ${result['usage']['billing_info']['total_cost']:.4f}")
print(f"Execution time: {result['execution_time']:.1f}s")

# Three additional target briefs from your pipeline
target_brief_2 = "Target: BlueShore Foods... [your second target]"
target_brief_3 = "Target: Helix Industrial... [your third target]"

briefs = [target_brief, target_brief_2, target_brief_3]

def make_task(brief: str) -> str:
    return (
        "Conduct a first-pass M&A diligence on the target below. "
        "Produce the deal memo per the swarm spec.\n\n"
        f"TARGET BRIEF:\n{brief}"
    )

batch_payload = [{**payload, "task": make_task(b)} for b in briefs]

batch_response = requests.post(
    f"{BASE_URL}/v1/swarm/batch/completions",
    headers=headers,
    json=batch_payload,
    timeout=1800,
)

for i, deal_result in enumerate(batch_response.json()):
    print(f"\n=== TARGET {i + 1} ===")
    print(
        f"Cost: ${deal_result['usage']['billing_info']['total_cost']:.4f} | "
        f"Status: {deal_result['status']}"
    )