Meet Melville.
Your due diligence analyst.Your due diligence analyst.
He already read the deck, the cap table, and the 2019 blog post.
Before you asked.
Create Account
Already have an account? Sign in
Find Risks, Conflicts, Opportunities.
Growth leans on manual data entry to beat hostile vendor APIs, so it won't replicate city to city without linear overhead. The venture-scale valuation rests on an assumption that doesn't hold.
The €560K of "ARR" is gross billings, not net revenue. At a 12.4% take rate the real figure is a fraction of that, and the €8M cap implies a multiple no institutional investor will accept.
The "no direct B2C competition" claim broke when a well-capitalized incumbent launched a consumer offering in early 2026. It pressures customer acquisition cost and premium supply at once.
Swap the legally toxic cash fine for forfeiting a pre-paid credit. Same penalty for no-shows, zero chargeback exposure.
Weeks of team hours.
Minutes of yours.
You send
- pitch-deck.pdf
- cap-table.xlsx
- financials.pdf
You brief
He reads
He reports
Think Claude can handle this?
Feed a model more and it starts to slip. Push a full data room through even the best one and it misses things, mixes them up, and makes them up. At report scale, as many as 4 in 10 facts it hands back were never in the file.
Model accuracy at 750 pages of context
GPT-5.6 Sol 61.3%
Claude Opus 4.6 54.3%
Gemini 3.5 Flash 38.8%
GLM 5.2 34.9%
DeepSeek V4 Pro 28.8%
500K tokens · 8-needle retrieval · benchmark from Context Arena
You + Frontier LLM
Melville
- Time to Due Diligence
- Your weekends. Plus a few evenings.
- Up to 3 hours
- Available formats
- Chat replies. You compose the memo.
- Report. Deck. Podcast. One-pager.
- Specialist perspectives
- As many as your patience allows
- 70+ specialist lenses
- Rounds of analysis
- Until you feel done
- 13 sequential rounds
- LLM interactions per analysis
- Until confidence sets in · usually ≤100
- ~1,500 per analysis
- Accuracy at scale
- Drops ~40% past 500K tokens
- 150M+ tokens per standard report
- Adversarial cross-check
- You be the skeptic. Manually.
- Two teams debate. Synthesis delivered.
- Source citation
- You verify. One by one.
- Every one, page-linked.
Wall-clock time to a defendable report.
What lands on the committee's desk.
Angles simulated on the same file, at once.
Times the file is re-read for what was missed.
Model calls behind a single answer.
How recall holds as the file grows.
Every finding argued both ways before you see it.
Every claim, traceable to a page.
Everywhere you work.
He never leaves. Not until exit. Pick how Melville shows up: an analyst in your channels, or an engine you wire into your own.
As an agent
He lives in Teams, Slack, and your inbox. He watches the file and the world. Every filing, board pack, and headline that could shift your risk, flagged before you ask. Drafted, source-linked, ready.
Slack
Teams
Outlook
Gmail
Melville MCP
Run Melville as an MCP server and call him from your own stack. Wire in Claude Code, Codex, or any MCP client, and pull diligence-grade research straight into the tools your team already builds with.
Claude Code
Codex
Gemini
+ any MCP client
Hire Melville All in one.
The whole file
Conflicts, omissions
Every claim cited
Before you sign
