Setup Guide

Connect lexicalpy to ChatGPT in five steps

Vocabulary NER and embedding-based semantic search, exposed as a Custom GPT action. The same server speaks MCP at /mcp/ for MCP clients.

Checking server
1

Get your API key

The server admin has set a bearer token. ChatGPT sends it on every call. If you don't have a key, ask your admin.

2

Create a Custom GPT

  1. Go to chat.openai.com/gpts/editor (Plus plan required).
  2. In the Configure tab, name your GPT and add a short description.
  3. Scroll to the bottom and click Create new action.
3

Import the OpenAPI schema

In the action editor, click Import from URL and paste this:

…

ChatGPT will load the schema and list three operations: ner, lookup, semantic_search.

4

Set authentication

  1. Click the gear icon next to Authentication.
  2. Choose API Key.
  3. Auth Type: Bearer.
  4. Paste your API key from Step 1.
5

Try it in your GPT

Save the GPT and send it a message like:

What MeSH terms match "opioid antagonist"? Use semantic_search.

The GPT calls /search-records and returns ranked vocabulary candidates with similarity scores.

Test the API directly

Paste your API key and a query to verify the endpoint before configuring ChatGPT.

Connect from other clients

The same server speaks MCP at …. Any client that supports Streamable HTTP transport can use it.

Codex CLI MCP

Edit ~/.codex/config.toml and add:

Save the file and restart Codex. Verify by asking it to call semantic_search for a vocabulary query. See Codex docs for the latest MCP config schema.