Skip to content

lohith261/keen

Repository files navigation

KEEN

AI that does your due diligence homework for you.

🌐 Try it live: keen-sigma.vercel.app — no account needed

Built at TinyFish Accelerator (2026 cohort)


So what even is this?

Imagine a private equity firm is thinking about buying a company. Before they write a big check, they need to dig into that company's finances, customers, contracts, software — basically everything. This process is called due diligence and it usually takes a team of analysts weeks to do manually.

KEEN does that job automatically.

You tell it the company name, click a button, and it goes off and reads financial systems, CRM databases, public filings, LinkedIn, and more — then comes back with a clear list of red flags, discrepancies, and things that need a second look. Like having a really fast analyst who never sleeps and doesn't charge by the hour.


Try it in 5 minutes (no sign-up needed)

Go to: keen-sigma.vercel.app

1. Check everything's working

Click STATUS in the top bar. If you see a green OPERATIONAL badge, you're good to go.

2. Open the Dashboard

Click DASHBOARD. No login needed — the app runs in Live mode by default, but you can connect your own data sources later if you want. For a quick look, just create a new engagement and let it run with the demo data.

3. Create an engagement

Click + New Engagement, type a company name (it'll auto-suggest from SEC filings as you type), and hit Start Pipeline.

4. Watch it work

Three AI agents run one after another — Research, Analysis, then Delivery. You can see exactly what each one is doing in real time, including a live view of the browser when it's navigating tools like LinkedIn or Bloomberg. The whole thing takes a couple of minutes.

You can pause, resume, or restart the pipeline at any point without losing progress.

5. Look at the findings

Once it's done, click the RESULTS tab. You'll see a list of things KEEN found — like:

  • Revenue numbers that don't match between two systems (a $1.3M gap)
  • A sales funnel that's leaking more leads than it should
  • An engineer who left the company that nobody flagged
  • Overdue customer invoices that are 4+ months old

Each engagement has its own unique URL — you can bookmark it, share it, or refresh without losing your place.

6. Export the report

Hit PDF or XLSX to download a proper report you could actually hand to someone. You can also push it straight to Google Sheets or Google Drive.


What it actually does under the hood

(Feel free to skip this bit — but here's a plain-English explanation if you're curious)

Step 1: Research

KEEN connects to a bunch of business software systems — things like Salesforce (where sales teams track deals), NetSuite (accounting software), SEC EDGAR (US government company filings), HubSpot (marketing data), and Crunchbase (startup funding info).

For systems that don't have a simple connection, it uses a tool called TinyFish that literally opens a web browser, logs in, and reads the data like a human would — except automatically. Think Bloomberg, LinkedIn, SAP, Oracle, and others. While this is happening, you can watch the live browser window directly in the dashboard.

It pulls all of this data together into one place.

Step 2: Analysis

An AI (Claude, made by Anthropic) then reads through all that data and looks for things that don't add up. If the sales system says revenue is $10M but the accounting system says $8.7M — that's a red flag. It flags these discrepancies, scores how serious they are, and marks the ones that need a human to double-check.

Step 3: Delivery

Finally, it writes up a proper report — executive summary, full findings, everything — and can automatically send it to Slack, email, Google Drive, or SharePoint.


What's built and working

Feature Status
Full 3-step AI pipeline (Research → Analysis → Delivery) ✅ Works
5 live data connections (Salesforce, NetSuite, SEC, HubSpot, Crunchbase) ✅ Works
Browser automation for locked-down systems (via TinyFish) ✅ Works
Live browser view — watch the AI navigate in real time ✅ Works
Pause, resume, and restart a pipeline mid-run ✅ Works
PDF and Excel report export ✅ Works
Push report to Google Sheets / Google Drive ✅ Works
Auto-send to Slack, email, SharePoint ✅ Works
Upload your own documents (PDF, Excel, Word, etc.) ✅ Works
Expert call transcripts (Tegus, Third Bridge + manual upload) ✅ Works
Deal comparison against similar deals (benchmarking) ✅ Works
Ongoing monitoring of a company after acquisition ✅ Works
Customer interviews, channel checks, win/loss analysis ✅ Works
External verification — court records, patents, bank statements ✅ Works
Technical review of a company's GitHub codebase ✅ Works
Virtual Data Room ingestion (Datasite, Intralinks) ✅ Works
Unique URL for each analysis — shareable and refresh-safe ✅ Works
Dark and light theme ✅ Works
Sign in / accounts (so your work is saved) ✅ Works
Encrypted vault for storing API keys and credentials ✅ Works
Legal contract clause scanning (CoC, IP, non-compete, litigation, regulatory) ✅ Works
Billing / pricing 🔲 Not built yet
Admin tools 🔲 Not built yet

Demo mode vs. Live mode

Demo Live
What you see Fake "Zendesk Inc" data Real data from your actual connected systems
Credentials needed None You'll need to connect your accounts
Good for Showing people how it works Actually using it for a real deal
Default No — app defaults to Live Yes — this is the default
How to switch Set keen-data-mode: demo in your browser's local storage Set keen-data-mode: live (or clear it)

Before you run in Live mode

When you first create an engagement in Live mode, you'll see a Credentials button. This is where you add the API keys or logins for each system you want KEEN to pull from. The credentials are encrypted and stored securely — they never leave your account.


Want to run it yourself locally?

You'll need a few things installed on your computer first:

  • Node.js (for the website part)
  • Python 3.11+ (for the AI engine)
  • Redis (a fast database, used for caching)
  • A free Supabase account (handles login)

Website (frontend)

cd frontend
npm install
cp .env.example .env.local    # add your Supabase details
npm run dev                    # opens at http://localhost:5173

AI engine (backend)

cd backend

python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

cp .env.example .env
# Fill in the details (database, Supabase, at least one AI key)

alembic upgrade head           # sets up the database tables
uvicorn app.main:app --reload --port 8000

Generate a secret key (you'll need this once)

python3 -c "import base64, os; print(base64.b64encode(os.urandom(32)).decode())"

Environment variables (the settings it needs)

Backend — required

Variable What it's for
DATABASE_URL Where the database lives
REDIS_URL Where the fast cache lives
SECRET_KEY A random secret the app uses internally
CREDENTIAL_ENCRYPTION_KEY Encrypts any API keys you enter
SUPABASE_URL Your Supabase project address
SUPABASE_ANON_KEY Supabase public key
CORS_ORIGINS Which websites are allowed to talk to the backend

AI keys — need at least one

Variable Which AI
ANTHROPIC_API_KEY Claude (this is the main one)
GEMINI_API_KEY Google Gemini (backup)
GROQ_API_KEY Groq (backup backup)

For live mode

Variable What it unlocks
TINYFISH_API_KEY Browser automation for Bloomberg, LinkedIn, SAP, Oracle, and more

Frontend — required

Variable What it's for
VITE_API_URL Where the backend is (e.g. https://keen-backend.fly.dev)
VITE_SUPABASE_URL Supabase project address
VITE_SUPABASE_ANON_KEY Supabase public key

How it's deployed (the live version)

  • Website → hosted on Vercel (auto-deploys whenever code is pushed to GitHub)
  • AI engine → hosted on Fly.io
  • Database → Supabase (managed, so we don't have to worry about it)
  • Cache → Redis on Fly.io

License

© 2026 KEEN — Backed by TinyFish Accelerator

About

AI-powered M&A due diligence with multi-agent orchestration

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors