Jeremy Longshore

On October 20, 2025, Anthropic announced Claude for Life Sciences - a suite of research tools for scientific literature. That same evening, I built something better: a production-ready PubMed research toolkit with 10 MCP tools, comprehensive test coverage, and zero security vulnerabilities.

This is how I approached the problem, the critical decisions I made, and why simpler solutions often win.

The Challenge: Build Better, Not Compete

When I saw Anthropic’s announcement, I didn’t think about competition. I thought about user value:

The answer: Build a complete toolkit that users can actually own and customize.

Problem-Solving Under Constraint

The first major challenge was technical: I needed to use Vertex AI Gemini 2.0 Flash for code generation, but it was trained before the Model Context Protocol existed.

The Wrong Approach

My initial attempt assumed the AI understood MCP. Result: Incomplete code, missing critical components, failed implementation.

The Right Approach

I created a 480-line context document explaining:

Lesson: When working with constraints, invest time in setup to save time in execution.

Second attempt with comprehensive context: Complete success.

Critical Design Decision: Simplicity vs. Features

The AI-generated code included SQLite caching - impressive, feature-rich, complex.

Then came user feedback: “What’s the SQLite database have to do with anything?”

This triggered a decision matrix:

OptionProsCons
Keep SQLiteOffline caching, comprehensive featuresComplex setup, more dependencies, harder to maintain
Remove SQLiteSimple, clean, easy to understandNo offline access by default

I chose simplicity. Here’s the business reasoning:

  1. Free version should be simple - Lower barrier to entry
  2. Premium upgrades for complexity - Monetization path
  3. User ownership - They can add features themselves (open source)

This decision reduced dependencies from 5 packages to 3, eliminated database management, and created a clear product differentiation strategy.

Technical Architecture: 10 Specialized Tools

Rather than one monolithic tool, I designed a modular system:

Research Discovery (Tools 1-4)

Network Expansion (Tools 5-7)

Analysis & Comparison (Tools 8-10)

Each tool has a single, well-defined purpose. No feature creep.

Quality Assurance: Test-Driven Validation

I built 16 comprehensive tests covering:

  1. Server initialization - Proper startup and configuration
  2. Tool registration - All 10 tools available
  3. Rate limiting - Actual timing tests with NCBI compliance
  4. Input validation - Parameter checking and error handling
  5. Error scenarios - Network failures, invalid inputs, API limits

Result: 16/16 tests passing, 0 security vulnerabilities, 0 TypeScript errors.

This wasn’t about checking boxes - it was about confidence in deployment.

Compliance & Best Practices

NCBI E-utilities has strict rate limits:

I implemented automatic enforcement with proper timing logic:

private async enforceRateLimit() {
  const now = Date.now();
  const timeSinceLastRequest = now - this.lastRequestTime;

  if (this.apiKey) {
    // 10 req/s = 100ms minimum delay
    if (timeSinceLastRequest < 100) {
      await new Promise(resolve =>
        setTimeout(resolve, 100 - timeSinceLastRequest)
      );
    }
  } else {
    // 3 req/s = track counter within 1-second window
    if (this.requestCounter >= 3 && timeSinceLastRequest < 1000) {
      await new Promise(resolve =>
        setTimeout(resolve, 1000 - timeSinceLastRequest)
      );
    }
    this.requestCounter = (timeSinceLastRequest < 1000)
      ? this.requestCounter + 1 : 1;
  }

  this.lastRequestTime = Date.now();
}

Every API call goes through rate limiting. No exceptions. Compliance isn’t optional.

Intelligent Automation: The 9.6 KB Agent Skill

While Anthropic uses 500-byte Agent Skills, I built a 9,600-byte Literature Review Automator (17x larger) that runs complete research workflows automatically.

When a user says “Review the literature on X,” it:

  1. Constructs optimized queries - Analyzes topic, identifies synonyms
  2. Retrieves comprehensive data - Articles, abstracts, metadata
  3. Analyzes patterns - Trends, key researchers, citation networks
  4. Synthesizes findings - Themed grouping, structured summaries

This isn’t just bigger - it’s systematically better because it guides Claude through a proven research methodology.

Product Positioning: Better Than Anthropic

I’m not claiming superiority out of ego. Here’s the objective comparison:

MetricAnthropicMy SolutionAdvantage
CostPaid tierFree (MIT)100% savings
Tools1 basic tool10 specialized tools10x functionality
Agent Skills500 bytes9,600 bytes17x more guidance
CustomizationProprietaryOpen sourceFull control
Data PrivacyCloud-hostedSelf-hostedUser ownership

This isn’t competition - it’s a different product philosophy.

Execution Timeline: 4 Hours

From concept to production in one evening.

What This Demonstrates

Technical Skills

Product Skills

Process Skills

Try It Yourself

The plugin is live in the Claude Code marketplace:

/plugin marketplace add jeremylongshore/claude-code-plugins
/plugin install pubmed-research-master@claude-code-plugins-plus

Or just ask Claude: “Review the literature on CRISPR gene editing”

The Source Code

Everything is open source:

Fork it. Examine it. Learn from it.

Building better solutions means understanding user needs, making thoughtful trade-offs, and executing with quality. This project demonstrates all three.

#Typescript #Mcp #Open-Source #Product-Development #Api-Design