If you're an active angel, your real bottleneck isn't capital. It's attention. You see decks, intros, cold inbounds, and warm handoffs every week, and the honest truth is that a lot of high-quality deals get dropped because you didn't have time to give them more than a glance. The deals that get a yes are usually the ones where the context happened to be loaded — you'd been thinking about that vertical, you'd seen a similar founder, you remembered the data point. The ones you missed are the ones where the context was loaded for someone else.
AI notes earn their keep here when they hold your funnel and your portfolio in one place where the agent can read across all of it, surface patterns, and pull forward the right context when a new deal hits the inbox. Not "ask AI to evaluate this deck." Ask the agent: how does this deck compare to the three closest things I've seen in the last year, what did I write about each of them, what's been the actual outcome on the ones I passed. The funnel-as-vault shape echoes AI notes for due diligence in acquisitions — the diligence side of the same picking discipline.
A deal-flow vault that mirrors how you actually work
The shape that holds up across angel styles is roughly: one inline database for the live pipeline, with rows for company, founder, source, sector, stage, status, and the link to the deal's notes page. One sub-page per deal that gets past the first screen, with the deck, the founder-call notes, the diligence summary, and any references. One cross-cutting page per portfolio company you've actually invested in, with the original thesis, the running founder-update file, and your evolving notes.
The pipeline database lives directly inside your main pipeline page via the :::database::: directive — you don't switch tabs to a separate tracker. The deal pages and portfolio pages are plain markdown notes the agent can read across in one query.
Capy supports unlimited page nesting, so you can let a long-running portfolio company fan out into a deep tree of update notes, while keeping a one-meeting pass as a single page. You don't have to commit to a structure in advance.
Screening that's actually grounded in your past screens
The hardest part of running a serious angel funnel is consistency. Two decks land in the same week, both look credible, and the difference between which one you take a call with and which one you pass on often comes down to which one happened to remind you of a deal you'd been thinking about. That's noise dressed up as taste.
The fix is to make the comparison explicit. Drop a new deck on a fresh page. Tell the agent: read this deck, then read every deck-screen note I've written in this sector over the last year, and tell me the three closest comparables. For each one, summarize what I wrote in the original screen, whether I took a call, whether I invested, and what the outcome has been. The agent reads the file and writes the comparison. You walk into your first reaction with the actual base rate of how this kind of deal has played for you, not just the most recent one you happened to remember.
PDF→markdown conversion via docstrange runs once on each deck you upload, so the agent can read the actual text — not just keywords from the file name — and surface the parts of the deck that match your comparables. The chat-with-PDFs mechanic at work — the same pattern explained in our writeup of Claude Code for documents, applied to deck screening.
Founder calls with speaker labels
Founder calls are the first real signal in a deal, and they're also the first thing to decay. The founder said something on the call about runway, or about the co-founder relationship, or about why the previous round was structured the way it was, and four weeks later when you're trying to remember whether to take the next call, you can recall the vibe but not the specifics.
Record the call inside Capy. The transcript comes back with speaker diarization — labels like "Speaker 1: …" so you can tell who said what. Park the recording on the deal's notes page. Ask the agent to draft a one-paragraph summary, pull every concrete number the founder cited, and flag anything that contradicts the deck.
The transcript also makes founder reference checks more useful. If two founders give different answers to a similar question across two calls, the agent can surface the contrast in a way scrolling through transcripts can't.
Diligence that doesn't restart from zero
Most angel diligence is light by nature — you're not a fund with a four-week diligence calendar — but the parts you do run tend to be the same shape across deals. Reference calls, market sizing, prior-round mechanics, founder background, customer or user pulse-check.
Tell the agent: build me a diligence checklist for this deal based on the diligence I ran on the closest comparable I've invested in, and pre-fill any rows that you can already answer from the deck or the founder-call transcript. The agent reads your past diligence pages and produces a checklist tailored to this deal — short on the things you've already verified, longer on the things that are open. You spend the saved time on the parts that genuinely need a human call.
The agent's web_search tool can pull current public information into the same diligence page when it's relevant — recent funding announcements, the company's current website copy, comparable-round data points — with source URLs so you can verify before you cite. You're still the one making the call; the agent is the research assistant.
Portfolio tracking that doesn't depend on you reading every update
The hard part of running an angel portfolio at scale isn't writing the original check. It's keeping context current on twenty or fifty companies whose monthly updates land in different formats, different cadences, and different inboxes. Most angels know which of their companies are in trouble within thirty days of finding out and are surprised by the rest.
A working setup: forward every founder update into the relevant company's page in Capy. Once a quarter, ask the agent to read across the updates for each company, summarize the trajectory — revenue trend, runway, hiring trend, key risks the founder named — and flag any company whose updates have shifted tone meaningfully since the last quarter. You walk into the quarterly review with a structured read instead of a vibes-based recollection.
The agent can also compare the most recent update against the original investment thesis you wrote when you wrote the check. If the company has materially drifted from the thesis, that's the kind of signal that's easy to miss in real time and obvious in retrospect — the agent surfacing it earlier is the small upgrade.
Pattern-matching across your own funnel
The compounding value of an angel notes layer is that, over time, your own funnel becomes a dataset you can ask questions of. "What's the win rate of cold inbounds vs. warm intros from a specific source over the last two years." "Which sectors have I passed on most often where the company went on to raise a real follow-on." "What did I write in the original screen for the deals that turned out to be my best outcomes — what pattern did I name."
Most of these questions are unanswerable today because the data lives across email, a spreadsheet, and a notebook. With the funnel and the portfolio in one vault, the agent can read across all of it and answer them. You're not running a fund-level analysis; you're checking your own pattern recognition against your actual outcomes. That's the kind of feedback loop that quietly makes angels better at picking.
What this isn't
Capy isn't a CRM or a portfolio-management system. It doesn't generate cap tables, calculate returns, or send templated update requests. The structured-data side of an angel practice still lives in the tools that handle structured data. Capy is for the unstructured side — the screen notes, the founder-call transcripts, the diligence files, the running portfolio commentary — which is the part of the day that's actually slow.
It's also single-user by design. One angel, one vault. If you syndicate with a group that wants a shared workspace with role-based access, that isn't this product. The shape that fits a personal angel practice is a vault that lives with the person making the calls.
A small first test
Pick one active deal. Load the deck, your notes from the founder call, and any references you've collected on a single page in Capy. Ask the agent to write a one-page summary of how this deal compares to the closest two things you've seen in the last year and what the strongest argument for and against investing is. If the summary names a tension you'd been carrying around without saying out loud, you've got a sense of what the agent does across the rest of the funnel. If you also run competitive intelligence — say, on portfolio-company competitors — the same vault holds that record too.
Try Docapybara free. Load one quarter of deal flow and see what the agent surfaces.