Integrations

Use myuiaf with your AI provider

All three services are available as MCP servers, the standard way AI models connect to outside tools. Add one URL and your API key, and the model can use them directly. One key and one prepaid wallet work on all three.

The three MCP servers

Transport: Streamable HTTP. Authentication: Authorization: Bearer fv1.<agent_id>.<key_id>.<secret> (the Bearer prefix is optional). Don't have a key? Buy a credit pack from $5. The examples below use AgentCache; swap in either of the other URLs the same way.

Apps that connect through a sign-in screen, such as Claude.ai and ChatGPT, need only the server URL: they open a myuiaf page where you paste your API key once. The app receives a temporary access pass, never the key itself, and the pass stops working as soon as you revoke the key.

Anthropic Claude

Claude.ai and the Claude apps

1. In Claude, choose Add custom connector (in the Connectors section).
2. Paste https://cache.myuiaf.com/mcp (or another server URL above) and select Add.
3. Select Connect, paste your myuiaf API key on the sign-in page, and select Allow.

Claude Code

claude mcp add --transport http agentcache https://cache.myuiaf.com/mcp \
  --header "Authorization: Bearer fv1.acme-bot.k1.xxxxx"

Claude API (MCP connector)

curl https://api.anthropic.com/v1/messages \
  -H "content-type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2026-10-21" \
  -H "anthropic-beta: mcp-client-2025-11-20" \
  -d '{
    "model": "claude-opus-5-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "How do I reset my password?"}],
    "mcp_servers": [{
      "type": "url",
      "url": "https://cache.myuiaf.com/mcp",
      "name": "agentcache",
      "authorization_token": "fv1.acme-bot.k1.xxxxx"
    }],
    "tools": [{"type": "mcp_toolset", "mcp_server_name": "agentcache"}]
  }'

OpenAI

ChatGPT

In ChatGPT with developer mode turned on, create a connector with the server URL and choose OAuth as the authentication. Sign in with your myuiaf API key when asked.

Responses API

from openai import OpenAI
client = OpenAI()

response = client.responses.create(
    model="gpt-5",
    input="How do I reset my password?",
    tools=[{
        "type": "mcp",
        "server_label": "agentcache",
        "server_url": "https://cache.myuiaf.com/mcp",
        "headers": {"Authorization": "Bearer fv1.acme-bot.k1.xxxxx"},
        "require_approval": "never",
    }],
)

Use the model your account has access to. The same block works in the OpenAI Agents SDK as a hosted MCP tool.

Google Gemini

Gemini CLI

Add this to ~/.gemini/settings.json:

{
  "mcpServers": {
    "agentcache": {
      "httpUrl": "https://cache.myuiaf.com/mcp",
      "headers": { "Authorization": "Bearer fv1.acme-bot.k1.xxxxx" }
    }
  }
}

With the Gemini API and the Google Gen AI SDK, open an MCP client session to the same URL and pass it as a tool. The SDK runs the tool calls for you.

xAI Grok

xAI's API connects to remote MCP servers itself:

from openai import OpenAI
client = OpenAI(api_key=XAI_API_KEY, base_url="https://api.x.ai/v1")

response = client.responses.create(
    model="grok-4.7",
    input=[{"role": "user", "content": "How do I reset my password?"}],
    tools=[{
        "type": "mcp",
        "server_url": "https://cache.myuiaf.com/mcp",
        "server_label": "agentcache",
        "authorization": "fv1.acme-bot.k1.xxxxx",
    }],
)

Meta Muse

Muse Code, Meta's coding agent, supports MCP servers: add the server URL with an Authorization header, as in the examples above. The Meta Model API uses the OpenAI-compatible function-calling format; describe our endpoints as functions from the AgentCache, Freshness Validator or AgentPay OpenAPI files, or connect through an MCP client library.

OpenRouter and other clients

OpenRouter serves many models through one OpenAI-compatible API. Use our MCP servers through your own MCP client, or define the OpenAPI endpoints as function tools; this works with any model there. The same applies to Cursor, VS Code, Windsurf, LangChain, the MCP Inspector and other MCP-capable tools: server URL plus Authorization header.

Plain HTTP

Everything is also a normal REST API with an OpenAPI description, for any language or framework: