ToolzerHub

For AI agents

Your agent asks.The public web answers.

One MCP server gives Claude, Cursor or any MCP client all 555 endpoints. Your agent finds the right one, calls it and reads the JSON.

Same key, same credits. Failed calls free.

  • Claude Code
  • Claude Desktop
  • Cursor
  • VS Code
  • Any MCP client

Install

Paste one command. Ask your first question.

Pick your client, paste the command, add your key. The three tools appear on the next prompt.

Terminal

claude mcp add --transport http toolzerhub https://api.toolzerhub.com/mcp \
  --header "x-api-key: your_api_key"

Find the most upvoted Reddit threads about standing desks this month.

search_endpoints"reddit search posts"
call_endpointGET /v1/reddit/search?q=standing+desk200

Here are the top threads, sorted by upvotes:

1
2
3
4
Sample interface data

Remote: https://api.toolzerhub.com/mcp over Streamable HTTP, key in x-api-key or Authorization: Bearer. Local: npx -y @toolzerhub/mcp with TOOLZERHUB_API_KEY.

How it works

Three tools. No glue code.

The server exposes three tools, so an agent goes from a plain question to a parsed result without you writing a client.

  1. 01

    search_endpoints

    Finds the endpoint for a job from a keyword, like "instagram reels".

    search_endpoints("tiktok hashtag videos")
  2. 02

    get_endpoint

    Returns its parameters, types, credit cost and an example request.

    get_endpoint("GET /v1/tiktok/hashtag_videos")
  3. 03

    call_endpoint

    Runs it with your key and returns the same JSON a normal call would.

    call_endpoint(id, { hashtag: "booktok" })

Try it

Ask in plain words. Get real data back.

Six prompts you can paste into Claude or Cursor once the server is installed.

  • Find the 20 most viewed TikTok videos under #skincare this week and list their creators.

  • Search the Meta Ad Library for ads mentioning "standing desk" and summarise the hooks they use.

  • Find this month's most upvoted Reddit threads about meal-prep containers and pull the top comments.

  • Compare the price and review count of these three eBay and Walmart listings in a table.

  • List 50 dental clinics in Austin from Google Maps with phone, rating and review count.

  • Pull every open data engineer job on Indeed in Berlin and group them by company.

Agent skill

One file that teaches the whole API.

SKILL.md in the agentskills.io format, from the public repo toolzerhub/agent-skills. Install it once and your coding agent knows how to call ToolzerHub without opening the docs.

  • The base URL and how to send the x-api-key header.
  • How to find the right endpoint and check its credit cost first.
  • The response shape: { data } or { error }, and what record_not_found means.
  • How to paginate and when a retryable error is worth retrying.

Terminal

npx skills add toolzerhub/agent-skills

Same credits, whichever way you call.

An agent call costs exactly what the same request costs from your code, and the docs list each endpoint's price. Failed calls cost 0 credits, and every account gets 100 free credits a month.

See pricing →

FAQ

Agent questions.

Still stuck? Ask us or read the docs.

What is the ToolzerHub MCP server?

The ToolzerHub MCP server gives AI agents access to 555 public web data endpoints through three tools: search_endpoints, get_endpoint and call_endpoint. It runs remotely at api.toolzerhub.com/mcp or locally with npx -y @toolzerhub/mcp.

Which agents and editors does it work with?

Claude Code, Claude Desktop, Cursor, VS Code and any client that speaks MCP over Streamable HTTP or stdio. The agent skill works with any coding agent that reads SKILL.md files.

Do agent calls cost more than normal API calls?

No. call_endpoint spends the same credits as a direct request, and failed calls cost 0 credits. Your agent can check an endpoint's price with get_endpoint before calling it.

Do I need a separate key for my agent?

No. The MCP server and the agent skill both use your normal ToolzerHub API key. Put it in the x-api-key header for the remote server or in TOOLZERHUB_API_KEY for the npx package.

MCP server or agent skill: which should I use?

Use the MCP server when you want the agent to call the API for you inside Claude or Cursor. Use the agent skill when the agent is writing code that calls ToolzerHub and needs to know the conventions.

Can I try it without paying?

Yes. Sign up free, get a key with 100 credits a month, and connect the MCP server. Most endpoints cost 1 credit, so that is up to 100 agent calls a month before you buy anything.

Give your agent a key.

Sign up free, paste one command, ask your first question.