A preclinical biotech runs their entire fundraise on a tool that didn't exist a week ago.

When the CFO reached out, fundraising was running on three disconnected systems:

• Their CRM for the contact graph
• Her inbox for the actual conversations
• Memory for who's hot, who's cold, who said what last week

Three sources of truth means none of them is right. She'd open the CRM to log a meeting, forget, and the activity log would silently rot. Worse — the team already had access to Houston, my biotech intelligence pipeline, which produces ranked daily signals about VC activity in their space. But there was no way to thread Houston's intelligence into her actual outreach. The signal showed up in one place; the action lived in another.

The CFO's job is decisions. Hers was being eaten by record-keeping.

This wasn't a "give us six months and a team" kind of build:

• Live in 4 weeks. Hard deadline.
• No new SaaS. The team had CRM fatigue. Whatever this was, they owned it.
• Claude has to be a first-class operator. Not a sidebar widget. Not a chatbot. A peer to the dashboard.
• Houston intelligence threaded into outreach. Not "open another tab" — actual integration.

What the CFO sees

• A custom-branded dashboard in her firm's exact color palette and logo, not mine.
• A daily Opportunities banner at the top of her homepage — Houston signals scored 7+/10, with paste-ready outreach copy already drafted.
• The pipeline — her firms, contacts, actions — all in one place, with the same status vocabulary she already uses ("waiting on me", "waiting on them", "parked").
• Claude as a co-equal operator. She can say to Claude.ai: "Log a meeting with the partner I just met with — he asked for the cap table" and Claude calls log_action against her database. The dashboard updates in real time.

What's under the hood

• 21 MCP tools covering full read/write of her pipeline (firms, contacts, actions, opportunities, aligned investors, aligned companies).
• Multi-tenant from day one — slug-based routing, per-tenant table prefixes, per-tenant brand theming via CSS variables. Onboarding tenant #2 is one focused day of work.
• Houston integration via a per-tenant linkage to my upstream intelligence pipeline. Her roster of aligned investors gets matched against fresh signals every morning.
• Auth — email + password, no magic-link friction.
• All hers. Code in her source control. Data in her database. Domain pointing to her infrastructure.

1. Talk to her pipeline.

That's the core unlock. Logging activity went from a task to a side effect of thinking out loud.

2. Outreach driven by intelligence.

The four "maybes" became zero.

• Time: 7 days end-to-end for v1, against a 4-week target.
• Money: Fixed-price build, scoped on the discovery call. Optional month-to-month retainer for the roadmap below. Pricing isn't public — get a number on a 30-minute call.
• Their commitment: A short discovery call, two weekly check-ins, no other meetings.

• Morning email digest — top opportunities arrive in her inbox before the day starts so she doesn't even need to open the dashboard.
• Outreach send — generate + send the model-drafted email directly from the opportunity card, logged automatically as an action.
• Investor signal alerts — push notifications when an aligned investor publishes anything significant.

The platform is multi-tenant. Onboarding the next biotech is roughly:

About 1 day of focused work end-to-end for the second client. The first one was the project; every one after is a checklist.