Skip to content

Getting Started

This tool accelerates M&A due diligence by analyzing your entire data room across 9 specialist domains (Legal, Finance, Commercial, ProductTech, Cybersecurity, HR, Tax, Regulatory, ESG) — helping your deal team find what gets buried, cross-reference it across domains, and trace every finding to an exact page and quote.

31% of M&A failures trace back to due diligence shortcomings, often because workstreams run in silos with no cross-referencing. This tool runs all nine workstreams simultaneously, cross-references findings across domains, and produces structured analysis your team can use as the foundation for IC memos, advisor reports, or negotiation checklists.

This tool does not replace professional advisors. Legal, financial, and regulatory conclusions should always be made by qualified professionals. This tool helps your team and advisors work more efficiently.

Prerequisites

  • Python 3.12 or later — check with python3 --version. If you need to install or upgrade, download from python.org.
  • An Anthropic API key — get one here. Alternatively, AWS Bedrock credentials work too.
  • A data room folder containing the contracts and documents to analyze.

Installation

pip install dd-agents[pdf]

This installs the tool and all required dependencies, including PDF extraction support.

Alternative: isolated install with pipx (recommended for CLI tools — avoids conflicts with other Python packages):

pipx install dd-agents[pdf]
Install from source (for development) 
git clone https://github.com/zoharbabin/due-diligence-agents.git
cd due-diligence-agents
pip install -e ".[dev,pdf]"

Optional Extras

Install these for additional capabilities:

pip install dd-agents[vector]     # Semantic search across documents (ChromaDB)
pip install dd-agents[ocr]        # OCR for scanned PDFs (English)
pip install dd-agents[glm-ocr]    # Multilingual OCR (100+ languages, Apple Silicon)

See When to Enable Vector Search below for guidance on dd-agents[vector].

Optional System Dependencies

Dependency macOS Linux Purpose
poppler brew install poppler apt install poppler-utils Fallback PDF extraction
tesseract brew install tesseract apt install tesseract-ocr OCR for scanned PDFs

These are optional — the tool works without them but may produce lower-quality text from some scanned documents.

API Key Setup

You need an API key to run the analysis. Choose one method:

Option A — .env file (recommended, persists across terminal sessions):

cp .env.example .env

Then edit .env and set your key:

ANTHROPIC_API_KEY=sk-ant-...

Option B — Environment variable (temporary, lasts until you close the terminal):

export ANTHROPIC_API_KEY="sk-ant-..."

Option C — AWS Bedrock:

export CLAUDE_CODE_USE_BEDROCK=1
export AWS_PROFILE=default
export AWS_REGION=us-east-1

Google Vertex AI and any other model (GPT, Gemini, …) via an Anthropic-compatible gateway are also supported — see Model Providers.

To override which AI model is used, pass --model-profile economy|standard|premium when running the pipeline (see Running the Pipeline).

Verify Installation

dd-agents version

This prints the installed version. If the command is not found, ensure the package installed correctly and your PATH includes the Python scripts directory.

Preparing Your Data Room

Organize your contracts into folders by subject (counterparty):

data_room/
  SubjectGroup_A/
    Acme_Corp/
      master_agreement.pdf
      amendment_2024.pdf
    Beta_Inc/
      license_agreement.pdf
  SubjectGroup_B/
    Gamma_LLC/
      services_contract.docx
  _reference/                    # Optional: reference docs (buyer overview, etc.)
    buyer_overview.pdf

Supported formats: PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), and images. Scanned PDFs are handled via OCR.

Folder structure matters: The tool uses folder names to identify which documents belong to which subject. A flat folder of files with no subfolder structure will still work — the tool groups them as a single entity — but organizing by subject produces better results.

A pre-built sample deal — Project Atlas — is included at examples/project-atlas/ so you can try the tool before setting up your own files. It is the same synthetic deal behind the live sample report. Run it with dd-agents run examples/project-atlas/deal-config.json.

Audio/Video Transcription

Audio and video files in the data room (.mp3, .wav, .m4a, .mp4, .mov, and similar) are automatically transcribed to text before analysis — no extra step needed. Board call recordings, earnings calls, or deposition audio drop into a subject folder like any other document.

Transcription tries three backends in order, using whichever is installed:

  1. mlx_whisper — Apple Silicon, fastest on macOS
  2. whisperx — GPU-accelerated, cross-platform
  3. openai-whisper — broadest compatibility, no GPU required

Install at least one (e.g. pip install openai-whisper) for transcription to run; without any backend, media files are skipped. Force a specific backend with DD_TRANSCRIPTION_BACKEND (mlx, whisperx, or openai), and override the model with DD_TRANSCRIPTION_MODEL.

Keyword search and agent tool access are sufficient for most data rooms. Vector search (pip install dd-agents[vector], backed by ChromaDB) adds value for semantic, cross-document queries, but isn't necessary for the core analysis pipeline below 500 documents — see Knowledge Architecture for the full reasoning. Past that threshold, or once cross-document semantic patterns become a primary analysis mode, install the extra.

If your data room crosses ~500 files and dd-agents[vector] isn't installed, dd-agents run logs an advisory warning (with --verbose) — it never blocks the pipeline.

Pre-Flight Check

Before running the full pipeline, assess your data room quality:

dd-agents assess ./data_room

This reports file type distribution, extraction readiness, and an overall completeness score. Address any critical issues before proceeding.

First Run

The typical workflow is three steps: generate a config, run the pipeline, review the report.

1. Generate a Deal Configuration

The fastest path is auto-config, which uses AI to scan your data room and produce a complete configuration:

dd-agents auto-config "Acme Corp" "Target Inc" --data-room ./data_room

This produces a deal-config.json with buyer/target details, company name variants, focus areas, and data room mapping. See Deal Configuration for details.

To preview the config without writing it:

dd-agents auto-config "Acme Corp" "Target Inc" --data-room ./data_room --dry-run

Alternatively, generate a config interactively without any API calls:

dd-agents init --data-room ./data_room

2. Run the Pipeline

dd-agents run deal-config.json

The pipeline extracts text, matches company names, runs AI analysis across all specialist domains, validates quality, and generates the report.

To preview what will happen without making API calls:

dd-agents run deal-config.json --dry-run

To also produce a red-flag stoplight triage signal alongside the full analysis:

dd-agents run deal-config.json --quick-scan --model-profile economy

See Running the Pipeline for all options including resume, model selection, and quality gates.

3. Review the Report

After the pipeline completes, find the outputs in _dd/forensic-dd/runs/latest/report/:

  • dd_report.html -- Interactive HTML report with cross-domain findings, severity filtering, and drill-down to exact clauses
  • dd_report.xlsx -- 16-sheet Excel report for detailed analysis and downstream work

Open dd_report.html in a browser. See Reading the Report for a walkthrough of each section.

Use these reports alongside your advisory process. The structured findings, citations, and cross-references serve as the foundation for your team's own deliverables — board presentations, advisor memos, negotiation checklists, or integration plans.

Post-Run Tools

Interactive Chat

Explore findings in a multi-turn conversation with document tools and persistent memory:

dd-agents chat --report _dd/forensic-dd/runs/latest

Ask follow-up questions, drill into source documents, verify citations — insights are saved automatically and recalled in future sessions.

Search contracts with custom questions without running the full pipeline:

dd-agents search prompts.json --data-room ./data_room

Natural Language Query

Ask a single question about findings (for multi-turn conversations, use chat instead):

dd-agents query --report _dd/forensic-dd/runs/latest -q "How many high-severity findings?"

PDF Export

Export the HTML report to a print-ready PDF:

dd-agents export-pdf _dd/forensic-dd/runs/latest/report/dd_report.html

Compare Runs

Diff findings between two arbitrary run directories — e.g. after new documents land, or to compare two analysts' runs:

dd-agents diff _dd/forensic-dd/runs/run_a _dd/forensic-dd/runs/run_b

Portfolio Management

Track multiple due diligence projects and compare risk profiles across deals:

dd-agents portfolio add "Alpha Acquisition" --data-room ./alpha_data_room
dd-agents portfolio list
dd-agents portfolio compare

Report Templates

Apply templates for different audiences (Board Summary, Legal Deep Dive, etc.):

dd-agents templates list
dd-agents templates show board_summary

See the CLI Reference for full documentation of all commands.

Docker

Build and run in a container:

docker build -t dd-agents .
docker run -e ANTHROPIC_API_KEY="sk-ant-..." \
  -v ./data_room:/workspace/data_room \
  -v ./deal-config.json:/workspace/deal-config.json \
  dd-agents run deal-config.json

Next Steps