500 free credits · 1 credit per successful fetch · Open-source MCP server
Server URL
https://mcp.unblockingapi.com/mcpcodex mcp add unblockingapi --url https://mcp.unblockingapi.com/mcpcodex mcp login unblockingapi; your browser opensThe app never sees your API key. Disconnect it any time under Connected apps.
Docs, product pages, search results and sites that block plain HTTP clients come back complete, not as a 403 or an empty shell.
For well-known sites a published template returns the fields - titles, prices, results - so Codex doesn't parse HTML.
Ask for a price monitor or a listings aggregator: Codex reads the build guide and writes an app that calls the API itself.
Codex can search the web, but reading a specific page that renders with JavaScript or sits behind bot protection is where agents stall. UnblockingAPI gives it a real browser fetcher: a plain fetch first, a browser when the site needs one, learned per site.
Each successful fetch costs 1 credit, failures are free, and new accounts get 500 free credits - enough to build and test an app before you buy a pack.
Paste the hosted URL and approve access. No key to copy, nothing to install.
codex mcp add unblockingapi --url https://mcp.unblockingapi.com/mcp
codex mcp login unblockingapilogin opens your browser: sign in to UnblockingAPI and allow access. The server is added to ~/.codex/config.toml, so it's in every project and in the Codex IDE extension.Same hosted URL, authenticated by your key in a header.
export UNBLOCKINGAPI_KEY=your_api_key_here
codex mcp add unblockingapi --url https://mcp.unblockingapi.com/mcp \
--bearer-token-env-var UNBLOCKINGAPI_KEY~/.codex/config.toml:[mcp_servers.unblockingapi]
url = "https://mcp.unblockingapi.com/mcp"
bearer_token_env_var = "UNBLOCKINGAPI_KEY"Prefer to run the server yourself? It is open source and runs with npx (Node.js 18+).
codex mcp add unblockingapi \
--env UNBLOCKINGAPI_KEY=your_api_key_here \
-- npx -y @unblockingapi/mcpAGENTS.md. Add this to the one in your project (or ~/.codex/AGENTS.md for every project) and it reads pages - and builds scrapers - through UnblockingAPI.## Web pages
When you need the contents of a URL, use the `unblock_fetch` tool from the
unblockingapi MCP server. Leave render unset: the API fetches plainly first and
switches to a real browser when the site needs one. Before parsing a well-known
site by hand, call `find_templates` with the URL - if a template covers it, pass
its reference as `unblock_fetch`'s template argument to get JSON.
When you write code that fetches web pages, call UnblockingAPI's REST API from
the backend (read the get_build_guide tool first), with the key in an
environment variable.find_templates and come back with parsed results instead of HTML.Use unblockingapi to get the listings on https://www.idealista.com/venta-viviendas/estepona-malaga/ as structured data, then summarise the three cheapest.Yes. The Codex CLI and the IDE extension read the same ~/.codex/config.toml, so the server you add with codex mcp add shows up in both.
Run codex mcp list: unblockingapi should be listed. Inside Codex, /mcp shows the server and its three tools. If sign-in expired, run codex mcp login unblockingapi again.
Yes - that's what the get_build_guide tool is for. Ask for "a price monitor using UnblockingAPI" and Codex reads the guide, picks templates and writes code that calls the REST API from your backend. See the prompt library.
Codex gets the page as Markdown or HTML and parses it itself.
Not when you sign in. Codex gets an access token for the hosted server, and the server uses your key on its behalf, so the key itself never leaves UnblockingAPI. You can disconnect any time under Connected apps. If you connect with an API key header instead, the key is stored in Codex's own config.
The MCP server is free and open source. Fetches run on your UnblockingAPI account: every successful fetch costs 1 credit and failed ones are free. New accounts get 500 free credits with no credit card, and bigger credit packs add bonus credits. Looking up templates is free.
Three, all read-only. unblock_fetch reads any URL through our proxy network (residential and mobile connections where a site needs them) and a real browser, picking a plain fetch or a browser per site by itself. It returns the page as Markdown (HTML on request), or JSON with a template. find_templates checks whether a published template covers a site. get_build_guide gives Codex the API reference and app guides when you build an app on the API. Details on the MCP server page.
The hosted server: nothing to install or update, and most apps sign in without a key. Run it locally with npx -y @unblockingapi/mcp if you prefer to keep the process on your machine - the tools and costs are the same.
One account, every assistant. Connect as many as you like, or see them all.