Created: 1/11/2026 8:31:47
Updated: 1/11/2026 9:21:53
Exported: 1/11/2026 9:22:47
Link: https://claude.ai/chat/0cc9b5d2-e022-48de-8ee8-1416f3a4688b
1/11/2026, 9:20:43 AM
I've done a lot of work on my medical knowledge graph project and pots rapidly approaching a point where I can build a corpus from 100 or 1000 papers and let anybody run sample queries against that corpus. And the next step would be something like an MCP server. I'm not sure from there how I turn it into a viable small business. Thoughts?
If it can pay me a workable salary, I'm absolutely happy to make it my job. As far as domain, I think initially work in oncology, since diagnosed there can be so heart breaking.
I don't touch patient data. I only do literature research with citations. Of somebody is doing a LLM session and happens to have patient data in a context window, I didn't put it there.
The contract with LexisNexis ended two months ago and I have limited runway. So I'm actively job hunting while I build. Assuming I get a job, work on this project will likely slow down. Though I'm getting much better at delegating to Gemini, Claude Code, and GitHub Copilot. I think I'm accidentally picking up some team leader skills.
Here's what I can find quickly online about SBIR loans.
You're eligible to apply for funding through America's Seed fund if your company:
- Is a for-profit entity located in the U.S.
- Has fewer than 500 employees (most applicants have fewer than 10).
- Is owned and controlled by U.S. citizens or permanent residents. Some additional restrictions apply with venture capital ownership.
By those criteria I'm good. I can't be non-profit, and that's OK.
You must have a Unique Entity ID (UEI) from SAM.gov to receive SBIR/STTR awards.
I don't have that yet. So far this is just a github repo and a bunch of little experiments.
1/11/2026, 9:21:53 AM
Thought process
Thought process: Architected hybrid business strategy combining MCP servers,
grant funding, and freemium models.
Will is at a really interesting juncture with his medical literature knowledge
graph project. Let me think through the business aspects here:
1. **Current state**: He's built a sophisticated medical knowledge graph system
that extracts entities, relationships, and evidence from research papers.
It's technically impressive - good schema design, proper provenance tracking,
uses modern tech (BioBERT, Pydantic, SQLModel, PostgreSQL/pgvector).
2. **Next steps he's considering**:
- Build a corpus from 100-1000 papers
- Create an MCP server to expose it
- Turn it into a viable business
3. **Constraints**:
- LexisNexis contract ended 2 months ago
- Limited runway
- Actively job hunting
- If he gets a job, development will slow down
4. **Business considerations**:
- Wants it to pay a workable salary
- Focusing on oncology initially (good choice - high stakes, lots of research, desperate need)
- Only literature research with citations (smart - avoids HIPAA/patient data issues)
- Considering SBIR funding
Let me think about viable business models:
**Option 1: SaaS for Clinicians/Researchers**
- Target: Oncologists, cancer researchers
- Value prop: "Literature-backed answers to clinical questions with full citations"
- Pricing: $99-299/month per user or institutional licenses
- Challenge: Sales cycle to hospitals/institutions is LONG
- Challenge: Needs to be really good to compete with UpToDate, PubMed, etc.
**Option 2: API/MCP Server for AI Companies**
- Target: Companies building medical AI assistants
- Value prop: "Add evidence-based medical knowledge to your LLM"
- Pricing: Usage-based API pricing or licensing
- Challenge: Competitive market, needs scale
**Option 3: Research Tool for Pharma/Biotech**
- Target: Drug development companies
- Value prop: "Competitive intelligence, target identification, mechanism insights"
- Pricing: Enterprise licenses $10K-100K+/year
- Challenge: Long sales cycles, need warm intros
**Option 4: Grant-Funded Research Project → Eventual Spinout**
- SBIR/STTR grants ($300K Phase I, $2M Phase II)
- Build it properly with government funding
- Challenge: Grant writing is a skill, takes time to get first award
Given his constraints, the grant route could actually work well - it buys time
and funding while he job hunts, and pharma/biotech licensing has higher deal
values than consumer SaaS. The key tension is that most paths require either a
strong sales network or significant product polish before they generate
revenue.
A phased approach makes sense here. First, secure employment for financial
stability while continuing to develop nights and weekends, with focus on
building a polished demo using real oncology examples and starting to gather
early interest. Then, launch an open-source MCP server to establish community
credibility and offer a $49-99/month hosted tier targeting individual
researchers and clinicians - this validates demand and generates early
traction. By month 6-12, that user base and feedback becomes a compelling SBIR
grant application, potentially unlocking $300K-2M in non-dilutive funding. The
MCP angle is particularly strong because it's currently trendy, solves a
genuine problem for LLM developers accessing medical knowledge, and positions
him in the developer market rather than competing directly with established
biotech platforms.
For monetization, open-source the protocol but charge for the hosted service,
premium curated corpora, higher-volume API access, and white-label deployments.
Start narrow - focus on one cancer type like breast cancer initially where
there's substantial research and clear clinical applications. A freemium model
works well: free tier with 100 monthly queries and 1,000 papers, Pro at
$49/month with unlimited queries and 10K papers, Enterprise with custom
pricing. Building publicly through technical blogs and sharing corpus insights
generates both credibility and organic user acquisition. Outreach to oncology
influencers and partnerships with medical institutions can accelerate adoption.
SBIR grants offer significant runway - get a UEI (free, 10 days) and apply to
NIH SBIR programs, particularly NCI's clinical decision support topics. Phase I
grants provide $300K over 6-12 months. Conservative revenue projections show
100 individual subscribers at $49/month plus 5 institutional licenses at
$500/month could generate roughly $90K annually, though a single SBIR Phase I
grant would fund proper development and Phase II applications worth $2M. The
main challenge is ensuring clinical accuracy and validation.
Looking at your med-lit-schema project, you've built something genuinely valuable — the architecture is solid, the provenance tracking is exactly what medical applications need, and the timing with MCP servers is perfect. Here's how I'd think about turning this into a business given your constraints:
With limited runway and active job hunting, you need a hybrid strategy: stable income + nights/weekends building. The good news: your "accidentally picking up team leader skills" with Claude/Gemini/Copilot is actually the key to making this work with limited time.
Your instinct about MCP is spot-on. Here's why it's a strong angle:
Open source the MCP server, monetize the corpus + infrastructure:
- Free tier: MCP server code (GitHub), basic oncology corpus (1,000 curated papers)
- Pro tier: $49-99/month - hosted service, 10K+ paper corpus, priority support
- Enterprise: Custom pricing - private corpora, white-label, API access
This works because:
- Developers want structured medical knowledge for AI apps
- Clinicians want literature-backed answers without technical setup
- You avoid the "UpToDate competitor" trap — you're infrastructure, not a product
Given your situation, apply for NIH SBIR grants while working:
Why this fits:
- Phase I: $300K over 6-12 months (that's your runway)
- Phase II: Up to $2M (that's quit-your-job money)
- Your demo + early users = strong application
- NCI (National Cancer Institute) specifically funds clinical decision support tools
Immediate steps:
- Get your UEI from SAM.gov (~10 days, free) - do this NOW
- Watch for SBIR solicitations (NIH publishes 3x/year)
- Your current corpus + MCP server = perfect Phase I deliverable
Grant angle: "Evidence-based clinical decision support infrastructure reducing time-to-insight for oncologists through AI-accessible knowledge graphs with full provenance tracking"
Months 0-3 (while job hunting):
- Build compelling oncology demo (pick ONE cancer type - breast cancer has great data)
- Launch basic MCP server as open source
- Start email list / early interest from oncology community
- Get UEI, research relevant SBIR topics
Months 3-6 (assuming you get a job):
- Nights/weekends: Launch hosted MCP service at $49/month
- Target: Oncology fellows, clinical researchers (they'll pay out-of-pocket)
- Goal: 20-50 early users = validation + testimonials for grant
- Submit first SBIR application
Your advantages:
- No patient data = no HIPAA nightmare
- Citations/provenance = clinically credible
- MCP timing = differentiated from existing tools
- Deep technical skills = can delegate to AI effectively
Revenue potential:
- 100 individual users × $49 = $58K/year
- 5 small institutions × $500 = $30K/year
- OR: 1 SBIR Phase I = $300K over 12 months
- Clinical accuracy: Partner with an oncologist advisor early (offer free access)
- Liability: Clear disclaimers — this is research infrastructure, not clinical advice
- Corpus quality: Start narrow (breast cancer), go deep, expand later
Heartbreaking diagnoses = desperate need for better tools. Specific niches to consider:
- BRCA-related cancers (your Olaparib examples suggest you're already thinking here)
- Rare oncology subtypes (where UpToDate is thin)
- Treatment resistance mechanisms (active research area, poor tooling)
If you pursue this:
- Form an LLC now ($100-200) - protects personal assets
- Open separate business bank account
- Track ALL expenses (deductible if this becomes real)
- Get UEI even if grants feel far away
You're at the "demo or die" stage. Don't try to build a complete business — build the smallest thing that someone will pay $49/month for. An MCP server with 1,000 well-curated breast cancer papers that answers "What's the evidence for Olaparib in BRCA1-mutated breast cancer?" better than PubMed searching would be worth money to the right people.
The SBIR path gives you runway without customers, but having even 10-20 paying users makes your grant application infinitely stronger.
Next concrete action: Get your SAM.gov UEI this week. It's free, takes 10 days, and keeps options open. Everything else can wait until you have income stability.
What aspect of this would be most helpful to dig into?