GRAPHGRAPHUse your @thunder.vc Google account.
What is waiting on you, in the order it deserves attention.
Drop in a company + the intent + any context. Thunder Graph reads it, finds the warmest, most relevant people in your network, and drafts the intro.
Ask anything about your people and firms — location, stage, sectors, check size, deal types, thesis. Warm relationships first.
Every company you've worked an intro for. Open one to track firms through the pipeline.
Run a mandate end to end: convert a scan into a client (economics + supporting data), curate the target list, then track every investor through the funnel.
Every mandate in one view: funnel health, fee pipeline, overdue actions, and conversion analytics that feed back into scoring.
What actually happened to the emails we sent. The Delivery Review agent reads each sending mailbox's bounce notifications every night, marks dead addresses, cancels queued sends to them, and has the Contact Sourcing agent find a verified replacement at the same firm.
The accounts Thunder Graph sends from on your behalf. Email runs on your Google Workspace mailbox and needs no setup. LinkedIn connects through Unipile — you sign in on their page, so your password and 2FA are never stored here. If a connection breaks, LinkedIn campaign steps are paused, not cancelled, and resume by themselves once it's healthy again.
Connect another Google account — a personal Gmail or a previous employer's Workspace — to backfill who you know into the graph. We read your calendar and contacts only: never the contents of any message. It adds and enriches contacts and firms and improves warmth scores. You can disconnect at any time, here or from your Google account.
Your signature — appended to every email Thunder Graph sends on your behalf, after the drafted message and after the AI writing check, so it is never rewritten, paraphrased, or dropped. Use Insert link for any URL: a link you insert this way shows as a clickable word to the recipient. Plain-text email clients cannot hide a web address behind a word, so the plain-text version of your emails shows the link's text only, with no address — never a written-out URL.
Everything the platform is doing, and everyone using it. Owner-only. Nothing here changes any data — it reads the same tables the app writes, so what you see is what actually happened.
Every AI prompt in the system, documented like API reference. Edit a prompt to override the code default (revert any time), test it with sample data, and see the log of its real runs.
Everything arriving at the top of the funnel. Website leads are thunder.vc form submissions — each one creates a contact and a firm/company record automatically, citing the form, unless it looked like spam or had no domain to anchor a firm on (those wait in Data review). Calendly bookings are inbound meetings. Prospects are companies scored as likely to need a $5M–$100M transaction soon — warm path, raise-window, news intent, SEC Form D, podcast guests. Convert one and it flows through the same Deal Room.
Your latest meetings and calls — expand a row to rate it and jot follow-ups (Claude pre-fills from the Fireflies transcript). Saving opens a reconnection thread the agent watches for the right moment to reach out.
Who to reach out to now and why — with a draft you can edit right here and send. Click a name or firm to open their full record. The agent watches your network (cadence, deal fit, news, your calendar) for the right moment, and learns from the reason you give when you dismiss someone.
Private events as a product — dinners, breakfasts, roundtables, receptions: set the city, theme and ideal attendee, link the clients it should serve, and Graph builds the invite list — warm first, location-gated — then runs the outreach and tracks RSVPs against Luma.
Autonomous sub-agents that work your network in the background. Each one shows exactly what it does, when it runs, the tools it uses, what it produces, and the prompt it follows — and you can run any of them on demand.
Records whose type looks wrong. Between Oct 2025 and Jun 2026 a bulk import wrote 1,687 rows into the production investor table mixing real firms with operating companies and individual people, then had an AI invent profiles for all of them — the firm types it produced were essentially random, which is why unrelated companies were showing as Family Offices. Each card shows the evidence. Your answer is final: nothing re-classifies a record you've decided on.
What's new across the firms in your network — fresh fundraises, deals, and news, refreshed in the background.
Every firm in the graph. Search, filter, open a 360° record, correct the data, or request AI enrichment. Edits persist through the nightly rebuild.
Every person in the graph — who they are, their persona, where they are, which Thunder owner knows them, and how warm. Filter, build tagged lists, export, or open a 360° record.