Where should I put the opencode.json config?

Use ~/.config/opencode/opencode.json for a global config that applies to every directory, or opencode.json at the project root for a per-project setup that you can commit to version control.

Why does OpenCode not recognise the crawler-sh server?

Make sure you used an absolute path in the command array. OpenCode does not expand ~ or environment variables. Also confirm the binary is executable with chmod +x ~/.crawler/bin/crawler-mcp.

Can I pass environment variables like CRAWLER_TOKEN in opencode.json?

Yes. Add an environment object inside the crawler-sh block. The exact key name may vary between OpenCode releases, so check the OpenCode MCP docs if a key is rejected.

How to Crawl a Website with OpenCode

OpenCode is a terminal agent that supports MCP servers. Once crawler-mcp is registered, you can ask OpenCode to crawl websites, fetch pages, and discover links - all from the command line.

This guide shows how to set it up and start crawling.

Step 1: Install crawler-mcp

Run the install script:

curl -fsSL https://install.crawler.sh/install-mcp.sh | sh

This downloads the correct binary for your platform to ~/.crawler/bin/crawler-mcp.

For more detail, see the installation guide.

Step 2: Wire it into OpenCode

Edit ~/.config/opencode/opencode.json (global) or opencode.json (per-project):

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "crawler-sh": {
      "type": "local",
      "command": ["/Users/you/.crawler/bin/crawler-mcp"],
      "enabled": true
    }
  }
}

Use the absolute path from which crawler-mcp. Restart OpenCode for the change to take effect.

For full setup details, see the OpenCode docs.

Step 3: Crawl a website

In OpenCode, ask:

Use crawler-sh to crawl https://example.com to depth 2 and list the page titles.

OpenCode calls crawl_site and returns the results.

Step 4: Fetch a single page

For one-off lookups:

Use crawler-sh to fetch https://example.com/blog/latest-post and summarise it.

OpenCode calls fetch_page and reasons over the Markdown response.

Step 5: Discover links

To map a site’s structure before a deep crawl:

Use crawler-sh to discover_links on https://example.com to depth 1.

OpenCode returns URLs with titles so you can decide which sections to explore.