CONNECT YOUR AI

SearchMesh + OpenAI API

Free during beta: 1,000 successful tool calls per account each UTC month. Failed calls are not counted.

1. Generate a SearchMesh credential below. It is not your OpenAI API key.
2. Set OPENAI_API_KEY and SEARCHMESH_API_KEY in server-side environment variables. Select a model available to your OpenAI account.
3. Add the remote MCP tool to a Responses API request. Keep approval enabled and handle approval requests before continuing. OpenAI API usage has separate costs.

MCP server URL

This client authenticates with a credential. Sign in to your dashboard, open OpenAI API under Connect your AI, and generate a key there. Keys are shown once, stored as hashes, and can be revoked individually. Send it as Authorization: Bearer YOUR_SEARCHMESH_API_KEY (or X-Api-Key).

Configuration / example

# Install: pip install openai
import os
from openai import OpenAI

response = OpenAI().responses.create(
    model=os.environ["OPENAI_MODEL"],
    tools=[{
        "type": "mcp", "server_label": "searchmesh",
        "server_url": "https://api.searchmesh.io/mcp",
        "authorization": os.environ["SEARCHMESH_API_KEY"],
        "require_approval": "always"
    }],
    input="Search for the official MCP specification and cite source URLs."
)
# Review mcp_approval_request items before sending mcp_approval_response.
# Approval policy remains your application's responsibility.
print(response.output)

Provider reference ↗

Try your first search

Ask your agent: Use SearchMesh to find the official MCP specification, read the relevant page, and cite the source URLs. Then check Request history in your dashboard.

Server authentication is supported; setup has not been tested end to end in every provider UI. Search queries and URLs are processed by SearchMesh and its upstream services; see the privacy policy.