Skip to main content
Zenrows gives your terminal, and your AI agents, reliable access to the protected web: sites that are dynamic, JavaScript-heavy, or built to block ordinary tools. The zenrows CLI puts fetching, extracting, crawling, and running jobs at scale behind one command, plus the skills, templates, and MCP config that make it work naturally with an agent, not just a script. It’s open source.

Installation

No API key needed to start. The first cloud call auto-provisions a Free-plan account and stores the key in .zenrows/secrets.json. Claim it later to keep your usage history, the same key keeps working. zenrows account claim opens the claim link, asking for a fresh one when the stored link is missing or 25 days old or more, and tells you if the account is already claimed:
To use an existing key instead of the auto-provisioning flow:

Run the CLI in CI

The CLI keeps its key in the project’s .zenrows/secrets.json, which a fresh CI checkout doesn’t have. Without a key, every run creates a new Free account, and the fourth run from the same network within 24 hours is refused with SIGNUP_RATE_LIMITED. Store your API key as a CI secret and pass it as ZENROWS_API_KEY, or cache the project’s .zenrows/ folder between runs. Add --no-signup to CI commands so a missing secret fails the job instead of creating an account. See Agent signup and unclaimed accounts.

Quickstart

Full setup

One command instead of many: creates your workspace, installs every skill, template, workflow, recipe, and eval, prints the MCP config to paste into your agent clients, and checks the capability matrix.
Target specific agent clients instead of the defaults:
init never skips a client silently. By default it targets claude-code, cursor, and vscode, and for each client it prints the config and the file it belongs in, where the client has one.

Commands

Primitives

The actual scraping work.

Agent Toolkit

What makes Zenrows easy for an agent, or a developer, to actually use. Check your setup, and what’s already installed:

Next steps

Connect an agent

Get the MCP config for Claude Code, Cursor, Windsurf, and more, printed by one command.

Troubleshooting

Read a failed run’s trace and find the specific next action.