Skip to main content
The AISA Model Context Protocol (MCP) server enables AI-powered code editors like Cursor and Windsurf, plus general-purpose tools like Claude Desktop, to interact directly with your AISA API and documentation.

​
What is MCP?

Model Context Protocol (MCP) is an open standard that allows AI applications to securely access external data sources and tools. The AISA MCP server provides AI agents with:
  • Direct API access to AISA functionality
  • Documentation search capabilities
  • Real-time data from your AISA account
  • Code generation assistance for AISA integrations

​
AISA MCP Server Setup

AISA hosts a remote MCP server at https://docs.aisa.one/mcp. Configure your AI development tools to connect to this server. If your APIs require authentication, you can pass in headers via query parameters or however headers are configured in your MCP client.
Add to ~/.cursor/mcp.json:
{
  "mcpServers": {
    "aisa-pdqs": {
      "url": "https://docs.aisa.one/mcp"
    }
  }
}

​
Testing Your MCP Setup

Once configured, restart your AI editor so it picks up the new MCP config, then try one of the following prompts in a fresh chat.

​
Example 1 — Discover an AIsa endpoint

Ask Cursor: “Using the AIsa MCP tools, find the endpoint for searching Polymarket events and show me a curl example.”
The agent calls the MCP search_docs tool, retrieves /api-reference/prediction-market/get_polymarket-events, and produces: 
curl "https://api.aisa.one/apis/v1/polymarket/events" \
  -H "Authorization: Bearer $AISA_API_KEY"

​
Example 2 — Generate a working integration

Ask Windsurf: “Write a Python function that uses AIsa’s Chat Completions API to summarize the text I’ll paste. Use the streaming format.”
The agent pulls the reference for POST /v1/chat/completions via MCP and generates: 
from openai import OpenAI

client = OpenAI(base_url="https://api.aisa.one/v1", api_key="...")

def summarize_stream(text: str, model: str = "gpt-5"):
    stream = client.chat.completions.create(
        model=model,
        messages=[
            {"role": "system", "content": "Summarize the user's text in 3 bullet points."},
            {"role": "user", "content": text},
        ],
        stream=True,
    )
    for chunk in stream:
        yield chunk.choices[0].delta.content or ""

​
Example 3 — Compare models for a task

Ask Claude Desktop: “Which AIsa model is cheapest for long-context code review, and what’s the per-1M input price?”
The agent queries the MCP list_models tool, filters by context window and coding capability, and returns:
gemini-2.5-flash-lite — 0.07/1Minputtokens,1,048,576−tokencontextwindow.Nextcheapest:‘qwen−flash‘at0.07 / 1M input tokens, 1,048,576-token context window. Next cheapest: `qwen-flash` at 0.022 with 1M tokens.

​
Available MCP tools

The AIsa MCP server exposes three tools to connected agents:
ToolWhat it does
search_docsFull-text search across guides, API reference, and skill docs
get_pageFetch a specific doc page by URL for detailed context
list_modelsReturn the model catalog with pricing and capability metadata
Your editor’s chat surface invokes these automatically when the agent determines they’re needed — you don’t call them by name.

​
Troubleshooting

Fix: Restart the editor after editing the config. Cursor/Windsurf load MCP servers at startup; changes don’t hot-reload.
Fix: Confirm your machine can reach https://docs.aisa.one/mcp — some corporate networks block non-standard paths. Retry; the MCP protocol is over HTTP and transient errors are safe to retry.
Fix: Different editors expose tools differently. If the agent can’t see the server, try prompting explicitly: “Use the aisa-pdqs MCP tools to find…”. The server name in your config (here aisa-pdqs) is the handle the agent needs to invoke it.

Model Catalog

What the list_models MCP tool returns.

API Reference

Endpoints the agent will reach through generated code.

Authentication

Bearer-token auth for generated API calls.