MCP server
The MCP server gives an AI client the catalog as tools it can call. It uses the same account, credits and lookup history as the REST API.
Connect a client
Sign in through the browser.
Add 1lookup.dev as a custom connector in Claude and sign in through the browser.
https://mcp.1lookup.dev/mcp- Open the connector settings in Claude and choose to add a custom connector.
- Name it 1lookup and paste the connector URL.
- Approve the sign-in page that opens, then enable the connector in a new chat.
https://mcp.1lookup.dev/mcpConnection details
| Setting | Value | Notes |
|---|---|---|
| URL | https://mcp.1lookup.dev/mcp | A remote server. There is nothing to install. |
| Transport | Streamable HTTP | Supported by current MCP clients. |
| Authentication | OAuth 2.1 or an API key | OAuth where the client supports it, otherwise a Bearer key header. |
How an assistant uses it
The server exposes 8 tools. Three cover the most common checks directly. The rest let the model reach all 79 products without a tool for each one.
search_productsfinds the right product for a task and reports its input fields and credit price.run_lookupruns any product by its slug, with the same JSON body the REST endpoint takes.get_accountreturns your credit balance and recent usage, so the model can tell you what it spent.
See MCP tools for every tool and its arguments.
Authentication
Clients that support OAuth 2.1 open a sign-in page, you approve access, and the client receives its own revocable token. Clients without OAuth send an API key in an Authorization: Bearer header.
Credits and visibility
- A tool call costs exactly what the same REST call costs. Repeats within 7 days are served from cache at no charge, and failed calls are not charged.
- Each result includes the credits used, so the model can report what it spent.
- Every call is recorded in lookup history with the client that made it.
- An assistant spends from the same balance as your code. Your plan's credits and your pay-as-you-go limit are the ceiling. See credits and billing.
Things to ask
Once the server is connected, describe the job in plain language. For example:
- Is +1 512 555 0142 a mobile number, and is it on the Do Not Call list?
- Verify these 20 emails and tell me which ones will bounce.
- Look up the tech stack and domain age for northwind.example.
- How many credits do I have left, and what did I spend this week?
For agents reading these docs
A plain-text index of every guide and endpoint is published at /llms.txt. Each doc page can also be copied or opened as Markdown from the menu next to its title.
Next
Each client also has an integration page with recipes: