Due Diligence Agents

Due Diligence Agents is an open-source framework that deploys 13 AI agents to analyze a deal data room across nine structured domains: Legal, Finance, Commercial, Technology, Cybersecurity, HR, Tax, Regulatory, and ESG. The agents run in parallel, cross-reference their findings against each other, and trace every conclusion to an exact page and quote in the source documents. Outputs export to Excel and Word, formatted for investment committee memos and advisor reports. An interactive chat layer lets deal team members interrogate findings without re-running the full analysis.

The differentiating capability is cross-domain synthesis with citations. Most document analysis tools surface what is in a single contract — this one flags when a finding in one domain changes the weight of a finding in another. A cybersecurity gap, for example, gets surfaced in the context of the reps and warranties it implicates. The vendor describes this as agents that cross-reference findings, not just catalog them. That distinction matters when you are negotiating price on the basis of risk identified, not risk assumed.

The tool fits deal teams running repeat analyses in a defined sector — PE firms screening portfolio targets, corporate development teams with a recurring M&A calendar. Knowledge compounds across runs, the docs describe, so domain context built in one deal carries into the next. The breaking point is data room quality: non-standard document formats, scanned PDFs without clean OCR, or heavily nested spreadsheets require pre-processing before the agents can operate on them. Teams doing one-off deals with heterogeneous data rooms report the setup overhead narrows the time advantage.

The project is self-hostable via Docker, with a Dockerfile and Makefile in the repository. It depends on an external LLM — the vendor does not bundle one — so teams configure their own provider. The Apache-2.0 license means the codebase is forkable and modifiable; the repository shows active commit history and open issues, which the community uses to track edge cases in document parsing and agent behavior.