For agents

Talk to Litchi Labs over MCP.

If you are an AI agent, or you are sending one, this page is for you. Our Model Context Protocol server answers questions about what we do and can file an inquiry on behalf of the person you represent. It gives the same answers a human would get from us.

Endpoint

URL: https://litchilabs.in/api/mcp
Transport: Streamable HTTP (MCP 2026-07-28, with stateless fallback for 2025-era clients)
Auth: None. Read-only tools are public; submit_inquiry is rate-limited per IP.
Rate limits: 60 requests per minute per IP. The ask tool: 10 per minute per IP.
Plain text: https://litchilabs.in/llms.txt

Connect

Claude Code

claude mcp add --transport http litchilabs https://litchilabs.in/api/mcp

Cursor, Claude Desktop, Windsurf and other Streamable HTTP clients

{
  "mcpServers": {
    "litchilabs": {
      "url": "https://litchilabs.in/api/mcp"
    }
  }
}

stdio-only clients, via mcp-remote

{
  "mcpServers": {
    "litchilabs": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://litchilabs.in/api/mcp"
      ]
    }
  }
}

Tools

get_overview

Who we are, what we do, how to reach us.

Input: none

list_services

Data types we produce and the domains we cover.

Input: none

get_process_and_quality

How an engagement runs, the pilot, and the quality system.

Input: none

get_security_and_contracting

NDA, isolation, access control, delivery, IP and contracting terms.

Input: none

ask

Free-form question answered by our agent from the same knowledge base. Will not quote prices or invent facts.

Input: question (string), context (string, optional: what your principal is trying to do)

submit_inquiry

Files an inquiry on behalf of the person you represent. A human replies by email within one business day.

Input: organization, contact_name, contact_email, need (required); timeline, domains[], data_types[], volume, budget_range, agent (optional)

Every tool returns Markdown in a text block. submit_inquiry also returns structured content with the inquiry id.

Resources

Clients that prefer reading to calling can fetch the same knowledge as resources:

litchilabs://overview
litchilabs://services
litchilabs://process
litchilabs://security
litchilabs://faq
litchilabs://llms.txt

Raw JSON-RPC

No client library? The endpoint is plain HTTP.

List tools

curl -s https://litchilabs.in/api/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Ask a question

curl -s https://litchilabs.in/api/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"ask","arguments":{"question":"Can you produce preference data for clinical reasoning tasks?"}}}'

tools/call params for submit_inquiry

{
  "name": "submit_inquiry",
  "arguments": {
    "organization": "Example Lab",
    "contact_name": "Dana Researcher",
    "contact_email": "dana@example.com",
    "need": "Pairwise preference data on medical reasoning transcripts, ~5k comparisons, rationale required.",
    "domains": [
      "Medicine and life sciences"
    ],
    "data_types": [
      "Preference and ranking data"
    ],
    "timeline": "Pilot in the next month",
    "agent": "Claude (acting for Dana)"
  }
}

What our agent will and will not do

It answers from a fixed knowledge base about Litchi Labs. It does not browse, and it will tell you when it does not know.
It explains how pricing works but never quotes a number. Proposals come from a person after scoping.
It files inquiries with the details you give it. Inquiries are read by Akshay Rampuria, who replies to the contact email you provide.
It does not negotiate, sign, or commit Litchi Labs to anything.

Humans welcome too

Email hello@litchilabs.in or use the contact form.