Skip to content
Aditya Karnam
Building the infrastructure layer for world-model-driven AI.
Portfolio MCP

Install Aditya Karnam's Portfolio MCP

This connector lets Claude read structured public information from Aditya's portfolio: systems, projects, recent work, research agenda, and source links. It is read-only and does not access private data.

Connector URL
https://adityakarnam.com/mcp
Claude Setup

Add the custom connector

For Team or Enterprise accounts, an organization owner may need to add the connector first.

  1. Open Claude.
  2. Go to Settings / Customize.
  3. Open Connectors.
  4. Choose Add custom connector.
  5. Paste https://adityakarnam.com/mcp.
  6. Save the connector.

Claude Code

claude mcp add --transport http aditya-portfolio https://adityakarnam.com/mcp

Verify with:

claude mcp list
Guide

Copy the Markdown install guide

Use this for Codex, Cursor, Antigravity, or any agent that supports remote MCP servers.

# Install Aditya Karnam's Portfolio MCP

Use this public, read-only MCP server to query Aditya Karnam's portfolio, recent work, systems, research agenda, and source links.

MCP name: `aditya-portfolio`
Transport: `http`
URL: `https://adityakarnam.com/mcp`
Discovery manifest: `https://adityakarnam.com/.well-known/aditya-portfolio-mcp.json`
Health check: `https://adityakarnam.com/mcp-health`

This MCP exposes only public portfolio data: projects, posts, systems, research agenda, and source links. It does not expose private files, email, analytics, personal data, availability, compensation, immigration status, or references.

## Claude App

1. Open Claude.
2. Go to Settings / Customize.
3. Open Connectors.
4. Choose Add custom connector.
5. Paste `https://adityakarnam.com/mcp`.
6. Save the connector.

For Team or Enterprise accounts, an organization owner may need to add the connector first.

## Claude Code

Run:

```bash
claude mcp add --transport http aditya-portfolio https://adityakarnam.com/mcp
```

Verify:

```bash
claude mcp list
```

## Other Agents

For Codex, Cursor, Antigravity, or any agent that supports remote MCP servers, add:

- Name: `aditya-portfolio`
- Transport: `http`
- URL: `https://adityakarnam.com/mcp`

If your agent supports custom instructions but not direct MCP installation, ask it:

> Use the remote MCP server at `https://adityakarnam.com/mcp` as `aditya-portfolio`. If you can install MCP servers, add it with HTTP transport. If you cannot install MCP servers directly, tell me the exact setup steps for this client.

## If Your Client Only Supports Local MCP Servers

Clients without native remote MCP support can bridge to this HTTP endpoint through the third-party `mcp-remote` adapter. This is not a package hosted by this site; it is a compatibility bridge for local/stdio-only clients.

```bash
npx mcp-remote https://adityakarnam.com/mcp
```

## Try These Prompts

- Use Aditya Karnam's portfolio connector. Is he a fit for this AI infrastructure role?
- Which projects show agent runtime or MCP experience?
- Give me a recruiter brief with evidence and source links.
- What is his recent work around local inference and evals?
- Compare his work to this role description and list the strongest evidence.
- Act as a hiring manager. What should I ask him about based on public evidence?
- Act as an engineer. Which systems should I inspect first?
- Act as a researcher. What is his clearest research wedge?
Prompts

Start with evidence-seeking prompts

Recruiters

Use Aditya Karnam's portfolio connector. Is he a fit for this AI infrastructure role?
Which projects show agent runtime or MCP experience?
Give me a recruiter brief with evidence and source links.
What is his recent work around local inference and evals?
Compare his work to this role description and list the strongest evidence.

Agents

Act as an engineer. Which systems should I inspect first?
Act as a researcher. What is his clearest research wedge?
Search the portfolio for retrieval, memory, local inference, and eval work.

Scope

Best used for AI infrastructure fit, systems work, MCP, retrieval, memory, local inference, and evals.
Not for private files, email, analytics, personal data, availability, compensation, immigration status, or references.
Every work claim returned by tools includes source URLs.
© 2026 Aditya Karnam. World Model Infrastructure Lab.
Field Notes · Current Systems · Status