AISOTools MCP Server

Query the AISOTools directory — 2742 AI tools across 27 categories — directly from your AI assistant over the Model Context Protocol. Read-only, no account, no API key.

For makersBuilt an AI tool? Get it in front of the people reading this page.

I built one — list it →

Endpoint (streamable HTTP)

https://aisotools.com/api/mcp

Setup

Claude Code

claude mcp add --transport http aisotools https://aisotools.com/api/mcp

Claude Desktop, Cursor, Windsurf, Cline

Add this to the client's MCP config file:

{
  "mcpServers": {
    "aisotools": {
      "type": "http",
      "url": "https://aisotools.com/api/mcp"
    }
  }
}

Clients that only speak stdio

Bridge through mcp-remote:

{
  "mcpServers": {
    "aisotools": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://aisotools.com/api/mcp"]
    }
  }
}

Tools

search_ai_toolsquery?, category?, pricing?, limit?

Ranked search across the catalog. Exact name and slug matches outrank tag and description matches, so asking for one specific product returns that product. Every response reports `matched` (total hits), `returned`, and `truncated`, so a small limit never reads as a small catalog.

get_ai_tooltool

The full record for one tool: description, features, pros, cons, who it is best for, pricing tiers, and curated alternatives. Takes a slug or a display name.

list_ai_tool_categories—

Every category with its slug, subcategories, and live tool count. The slugs are what search_ai_tools filters on.

compare_ai_toolstools[2-5]

Side-by-side records for 2–5 named products. Names it could not resolve come back in an `unmatched` array rather than being dropped silently.

check_ai_visibilityproduct_name, url, category?

Checks whether ChatGPT recommends a product. It reads the product's homepage, asks ChatGPT five buying questions a customer in that category would ask, and returns which questions named the product, which rival products were named instead, and a 0–100 score. Takes 10–30 seconds; a check of the same product from the last 24 hours is reused. See how it works below.

find_ai_tool_alternativestool, limit?

Curated alternatives first, then same-category tools to fill the request — the two are returned in separate fields, not blended.

How the AI visibility check works

check_ai_visibility runs the same check as the free scan on this site. Given a product name and its website, it:

reads the product's homepage to work out what its buyers are shopping for (or uses the category you pass);
asks ChatGPT (OpenAI's gpt-4o-mini, through the API) five buying questions a customer in that category would ask, without naming the product;
reports, for each question, whether the product was named and which other products were named, plus the products named most often overall and a 0–100 score.

Limits. It is one run of one assistant. Answers vary between runs and between assistants, so treat a result as a sample rather than a ranking. Only public websites can be checked. A check of the same product in the last 24 hours is reused instead of asking again. Each connection can run 5 new checks an hour, and the server runs a capped number per day; past that it asks you to try again later.

What is stored. The product name, website, category, the questions and the answers' findings are stored so the 24-hour reuse works; no account, email or chat content is collected. See the privacy policy.

Claude connector

To add only the AI visibility check to Claude, use this URL as a custom connector (Settings → Connectors → Add custom connector):

https://aisotools.com/api/mcp/claude

Try it without a client

curl -s https://aisotools.com/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"search_ai_tools",
                 "arguments":{"query":"transcription","limit":3}}}'

The catalog answered that. What does it say about your tool?

If you ship an AI tool, run a free AI-visibility scan on it — we put 5 buying-intent prompts to ChatGPT and score how often you get named instead of a competitor. ~30 seconds, no signup, no card.

Scan my tool — free →

Limits worth knowing

Read-only. No tool changes anything on your side or in the catalog — submitting a tool still goes through the submission form. The visibility check stores its own result for 24-hour reuse.
Stateless, POST-only. No sessions, no SSE. GET and DELETE answer 405 with a JSON-RPC error explaining why.
Catalog, not a live crawler. Pricing and features are what the catalog records, refreshed on our schedule rather than on request. Check the vendor's own page before acting on a price.
Curated alternatives are hand-maintained and can be shorter than your requested limit; the remainder is filled from the same category and labelled sameCategory so you can weight it differently.

FAQ

Do I need an API key?

No. The endpoint is read-only, unauthenticated, and free. It exposes the same catalog data that is already public on aisotools.com.

Which MCP clients does it work with?

Any client that speaks streamable HTTP — Claude Code, Claude Desktop, Cursor, Windsurf, and Cline among them. Clients that only speak stdio can bridge to it with mcp-remote.

Does it support SSE or sessions?

No. The server is stateless: it issues no session id and holds no server-initiated stream, so GET and DELETE return 405 with an explanation. Every POST is a self-contained JSON-RPC request and answer.

How fresh is the data?

The catalog ships with the site build, so the MCP server and the website always return the same records. Entries carry addedDate, and pricing is the pricing shown on the tool's page.

Are sponsored listings marked?

Yes. Every result carries a `sponsored` boolean, and a paid placement whose sponsorship window has expired reports false rather than staying flagged. `verified` marks entries reviewed by hand.

Privacy and support

Built something that belongs in here? Submit it to the directory and every client wired to this endpoint can find it.