Model Context Protocol (MCP) Substrate

Serve Abhishek Kumar's verified credentials, production case studies, and engineering telemetry directly into Cursor, Claude Desktop, or autonomous agent workflows via the Model Context Protocol.

Production Server Endpoints & Protocols

Connect external LLM systems via JSON-RPC 2.0 over HTTP or run the Python FastMCP server.

HTTP JSON-RPC 2.0FastMCP 4.0

1. Live Hosted Endpoint (Zero Setup)

Online

Connect directly to the production endpoint without running any local servers:

https://abhishekkumar.ai/api/mcp

Supports initialize, tools/list, tools/call, and resources/read.

2. Local FastMCP Server (uv / python)

To run the FastMCP server in your local terminal:

uv run --with fastmcp python mcp_server.py --port 8000

Serves on http://127.0.0.1:8000/mcp using Starlette/Uvicorn.

Client Configuration (Claude Desktop & Cursor)

Add this entry into your claude_desktop_config.json or Cursor MCP settings:

{
  "mcpServers": {
    "abhishekkumar-portfolio": {
      "url": "https://abhishekkumar.ai/api/mcp"
    }
  }
}

Exposed Tools & Resources (9 Endpoints)

Each tool exposes deterministic, typed JSON schemas for reliable multi-agent deliberation and tool use.

get_profile()Identity & Contacts

Verified profile, title, positioning, contact channels, and links.

get_flagship_metrics()Telemetry & Impact

Verified production metrics: 60–70% GPU spend cut, 3M→3K HTS triage.

list_projects()Systems Portfolio

BMS Compound Library, Eli Lilly RAG, Autoscaled vLLM on EKS.

get_project_details()Architecture Blueprint

In-depth architecture specifications, problem framing, and stack.

get_career_history()Chronology

10-year trajectory across BMS, Lilly, Biofourmis, Maple Labs, Accenture.

get_skills_by_category()Technical Taxonomy

Categorized taxonomy across GenAI, Vector Search, Cloud/K8s, and Python.

list_technical_briefs()Engineering Notes

Retrospectives on GPU serving, multi-agent debate, and hybrid RAG.

query_candidate_fit()Candidate Evaluation

Evaluates requirements against skills; returns match score and rationale.

cv://resume.mdMCP Resource

Raw markdown document stream of Abhishek Kumar's complete CV.

Empirical Verification via cURL

Test the live JSON-RPC 2.0 tool execution from any command line:

curl -X POST https://abhishekkumar.ai/api/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_profile","arguments":{}}}'