AIsa-mcp-server

AIsa-mcp-server is an MCP (Model Context Protocol) server designed to provide AIsa's data APIs to AI agents in the form of tools. Its core design is: instead of directly exposing hundreds of tools, it uses five meta-tools to search, inspect, and invoke the entire catalog, thereby saving context window space.

Purpose

This server is used to let AI agents access a variety of data APIs, covering areas such as SEO and AI visibility, finance, social, web search, sales, and agent email. It is suitable for agent scenarios that require dynamic discovery and invocation of external data services, such as market research, competitive analysis, sales lead generation, and more.

Main Capabilities

• Provides 580+ tools, distributed across 26 servers.
• Access via five meta-tools: search (search tools), get_details (get tool details), list_categories (list categories), use (invoke a tool), batch_use (batch invoke tools).
• Supports Streamable HTTP transport, and can be connected with a single URL.
• Supports OAuth authorization, with no need to create an account in advance; it also supports authentication using an API key (Bearer Token).
• Provides the stdio bridge package @aisa-one/mcp for clients that only support stdio (such as Claude Desktop).
• Supports pinning by category, and each category path still retains the five meta-tools and a price cap.
• Each invocation can set max_price_usd; invocations that exceed the budget will be rejected, avoiding unexpected charges.

How to Use

1. For clients that support remote MCP (such as Cursor, VS Code, Windsurf), add the URL in the MCP settings:https://mcp.aisa.one/mcp, with transport set to streamable-http.
2. Authorization uses OAuth: unauthenticated calls will return 401 and a WWW-Authenticate header pointing to https://mcp.aisa.one/.well-known/oauth-protected-resource/mcp, and the client can automatically register and open a browser to complete authorization.
3. If you already have an API key (from console.aisa.one), you can send it in the request header: Authorization: Bearer $AISA_API_KEY, with no browser authorization required, suitable for SSH or CI environments.
4. For clients that only support stdio, use the package @aisa-one/mcp as a local command; it will bridge to the remote endpoint, run the OAuth browser flow on first use, and cache the token. The environment variable AISA_API_KEY will be automatically recognized.
5. The range of tools returned by tools/list can be limited by pinning category paths, but this does not affect accessibility.

Applicable Scenarios

• When an AI agent needs to dynamically discover and invoke multiple data APIs.
• When cost control is needed to prevent the agent from unexpectedly invoking high-cost tools.
• In environments where browser authorization cannot be used (such as SSH, CI), access via API key.
• When using an MCP client that only supports stdio (such as Claude Desktop).

Limitations

• Search, get details, and list categories are free; use and batch_use are billed per call, with pricing the same as the AIsa REST API, and no seat fees or monthly minimum spend.
• Calls may return a 402 error due to insufficient balance or lack of a Hive GTM Growth subscription.
• The number of tools and categories may change with the catalog; real-time information can be found at mcp.aisa.one/servers.
• This repository mainly provides a stdio bridge; most clients can directly use the remote URL without this package.
• For more information, see mcp.aisa.one/llms.txt、mcp.aisa.one/servers、mcp.aisa.one/.well-known/mcp/server-card.json and aisa.one/mcp。