Open source · Apache-2.0 · Python 3.12+

Legal flags a risk.
Finance flags another.
We connect both.

13 AI agents read your entire data room across 9 specialist domains, cross-reference the findings no single reviewer links, and trace every conclusion to an exact verbatim quote.

Get started See sample report → View deck →

$ pip install dd-agents

PyPI · dd-agents Apache-2.0 4,000+ tests Docker Hub Homebrew tap

See it run

38 steps.
Every finding cited.

Point it at a folder. Quality-gated HTML report and 16-sheet Excel export — every conclusion traced to page, section, and verbatim quote.

13

AI agents
in parallel

9

specialist
domains

16

Excel sheets
per report

dd-agents run deal-config.json

▶ Step 06/38 build_inventory Discovered 247 documents across 14 folders ▶ Step 12/38 run_specialists Legal · Finance · Commercial · ProductTech Cybersecurity · HR · Tax · Regulatory · ESG ▶ Step 18/38 cross_domain_analysis ⚡ trigger IP clause (Legal §4.2) × ARR projection (Finance) CRITICAL IP restriction threatens 38% of projected ARR "perpetual license excludes SaaS delivery" MSA p.7 §4.2 ▶ Step 27/38 merge_findings 64 findings — 3 CRITICAL · 11 HIGH · 28 MEDIUM ▶ Step 38/38 generate_report ✓ output/report.html (interactive HTML report) ✓ output/report.xlsx (16-sheet Excel export) ✓ Pipeline complete 4m 12s

Features

Built for the diligence,
not the demo.

No vendor lock-in. No per-seat pricing. Run inside your own infrastructure with your own LLM provider.

01

Cross-domain analysis

A legal clause that threatens a finance projection. A tech dependency that undermines a commercial term. Compound risks surface automatically — named, cited, and severity-ranked across all nine domains.

02

Every finding, cited to source

Each conclusion traces back to a page number, section heading, and verbatim quote. Verifiable in under 60 seconds. The pipeline halts rather than ship output it can't ground.

03

Your data stays in your stack

Documents travel only as API calls to your configured LLM provider — Anthropic, AWS Bedrock, Google Vertex, or any Anthropic-compatible gateway. Your infrastructure, your audit trail.

04

Structured, auditable output

Interactive HTML report with Go/No-Go view, 16-sheet Excel export, JSON findings index, and IC memo template — structured for your advisors to verify, annotate, and build on.

Why open source

Your firm's AI,
on your
own terms.

Apache-2.0. No per-seat pricing that rises with every deal. Run it on your own infrastructure or fork it to match your firm's exact workflow.

Zero license cost

Apache-2.0. No per-seat pricing that rises with every deal. You pay only your model API cost — use any provider you negotiate directly.

Self-hostable by design

Deploy inside your perimeter. Compliance, data residency, and privilege-review controls stay under your firm's control — not a vendor's SLA.

Fork it. Extend it.

Add practice-specific workflows, integrate your DMS, wire in your own citation engine. The full codebase is yours to read, audit, and shape.

Audit every line

No black boxes. Open code, open prompt engineering. Know exactly how citations are parsed, severity scored, and data flows end to end.

Start analyzing.
Or clone the repo.

Use it as-is, or deploy the source inside your own environment.

Read the docs GitHub