Due Diligence Agents CLI
Python CLI where 13 AI agents read an M&A data room across nine domains and report with citations checked against the source
- 136 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Reads data room documents and sends their fragments to the chosen model API, and deal documents are usually confidential
- Spends paid API tokens and writes reports and intermediate files to the _dd/ directory
Install
Manual install
pip install 'dd-agents[pdf]'Python 3.12 and an API key in ANTHROPIC_API_KEY are required. Then dd-agents auto-config, dd-agents doctor and dd-agents run deal-config.json.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Reads data room documents and sends their fragments to the chosen model API, and deal documents are usually confidential
- Spends paid API tokens and writes reports and intermediate files to the _dd/ directory
This is third-party code. Review the repository files before installing.
What it does
Due Diligence Agents runs a forensic legal and financial review of a deal from data room documents. Nine specialists read every document in parallel through their lenses: legal, finance, commercial, product and tech, cybersecurity, HR, tax, regulatory and ESG. Four orchestration agents synthesize findings: a judge adversarially spot-checks conclusions, an executive synthesis calibrates severity and gives a Go or No-Go signal, a red flag scanner does fast triage and an acquirer intelligence module maps findings to the buyer thesis. The agents read documents directly with no RAG retrieval, citations are checked against the extracted source text and key quotes get a deterministic check. The pipeline halts when five blocking quality gates fail and can resume from any step. The output is an interactive HTML report, a 16-sheet Excel file and per-subject JSON with citations and cross-references. Commands: dd-agents auto-config builds a deal config, dd-agents doctor verifies model routing, dd-agents run deal-config.json runs the full review, with --quick-scan a quick triage, and dd-agents search asks questions across all contracts through a plain JSON of prompts. After a run dd-agents chat and dd-agents query work.
Who it is for. Lawyers, investment analysts and corporate M&A teams who need a first-pass data room review with document citations before specialists step in.
Good fit when
- A data room of hundreds of documents needs a first pass through legal, financial and tax lenses
- Before a full analysis you need a quick GREEN, YELLOW and RED triage
- All contracts need the same question about consent on change of control, with citations
- The result is wanted as Excel and interactive HTML for a team discussion
- A run must resume from the middle after stopping at a quality check
Not a fit when
- You need a legal opinion: the tool assists analysis and does not replace a lawyer
- Documents must not be sent to a model through an API: files leave the machine as API calls to the chosen model
- You need exact OCR of scans without extra install: scans need the [ocr] or [glm-ocr] extras
- The task is not about deals: the specialists are tuned for due diligence domains
Example request
Run dd-agents auto-config "Buyer Corp" "Target Inc" --data-room ./data_room, check the config with doctor, then start a quick triage with the --quick-scan flagLimitations
- Python 3.12 or newer and an API key are required: Anthropic, or AWS Bedrock, Google Vertex AI or a compatible gateway; documents leave the machine there as API calls.
- PDF needs the [pdf] extra with pymupdf under AGPL-3.0, scans need [ocr] or [glm-ocr], and tesseract and poppler are installed separately.
- The output assists analysis and does not replace a lawyer. The pipeline halts on blocking quality failures and gives no unreliable result.
- The tool reads the data room read-only and writes results to a separate _dd/ directory.
How to disable. pip uninstall dd-agents removes the tool, and Homebrew and Docker remove the install with their own commands. Results sit in the _dd/ directory next to the data room and are deleted manually.
FAQ
How do I install Due Diligence Agents?
Run pip install 'dd-agents[pdf]'. Alternatives: brew install zoharbabin/due-diligence-agents/dd-agents or the Docker image zoharbabin/due-diligence-agents. Python 3.12 and an API key are required.
Which domains do the agents analyze?
Nine: legal, finance, commercial, product and tech, cybersecurity, HR, tax, regulatory and ESG. They are enabled and disabled in deal-config.json.
How does a quick scan differ from a full run?
dd-agents run deal-config.json --quick-scan --model-profile economy gives a GREEN, YELLOW or RED signal across eight deal-killer categories, while a full run produces a report with a verdict, Excel sheets and JSON.
Where do my documents go?
Analysis runs on your machine and documents leave it only as API calls to the chosen model: Anthropic, AWS Bedrock, Google Vertex AI or a compatible gateway. There is no telemetry.
Can I ask one question across all contracts?
Yes, dd-agents search prompts.json --data-room ./data_room builds an Excel file with answers, citations and verification scores. Any lawyer can write the prompts file in plain JSON.
Discussion
No questions or reviews yet.
Related
CLI toolsA set of Claude Code skills: contract review, risk scoring, NDA and policy generation, compliance checks and PDF reports
A plugin helps agents maintain project documentation, decisions and verification evidence across development sessions
A contract review skill: red flags, key terms, comparison with market practice and ready redline wording that accounts for your side of the deal
A Claude Code and Codex plugin: a managing partner, 17 specialist lawyers, 226 skills and 20 workflows for Indian, US and UK law
A Claude Code skill for scientific literature reviews: an interview, Scopus and OpenAlex search, DOI checks and a final docx file
AFDocs tests a documentation site for AI agent friendliness: 28 checks across seven categories and a score out of 100.
README in short
The page shows the project idea: legal, finance and other specialists flag different problems and the tool links and cites findings. Then come the modes (full pipeline, quick scan, contract search, post-run tools), a quick start, the 13 agents diagram, security and privacy, the report contents, install and the documentation list.