kagi-enrich

What This Skill Does

The kagi-enrich skill provides OpenClaw agents with direct access to Kagi's proprietary, high-quality search indexes: Teclis and TinyGem. Unlike general-purpose search engines that prioritize mainstream corporate media and SEO-optimized content, this tool focuses on the "small web." Teclis excels at surfacing personal blogs, open-source documentation, and independent creative projects. TinyGem is specifically engineered to surface non-mainstream news, niche journalistic insights, and organic community discussions that often disappear from standard search results. By leveraging these specialized indexes, your OpenClaw agent can gather research that is more authentic, less commercialized, and uniquely suited for deep-dive technical or intellectual exploration.

Installation

To integrate this skill, ensure you have the OpenClaw environment configured. Run the following command in your terminal: clawhub install openclaw/skills/skills/joelazar/kagi-enrich

Prerequisites:

1. Obtain an API key from your Kagi settings (Settings → Advanced → API portal).
2. Set the environment variable by adding export KAGI_API_KEY="your_key_here" to your shell profile (~/.zprofile or ~/.bashrc).
3. Ensure your Kagi billing account is funded, as usage is billed at $0.002 per successful query.

Use Cases

This skill is ideal for research tasks requiring high-signal, low-noise information. Use it for finding technical troubleshooting guides that aren't buried under advertisements, discovering independent software projects that aren't yet trending on social platforms, or conducting media studies that require perspectives outside the mainstream press echo chamber. It is an excellent supplement to traditional web searches when the agent needs to find "the human side" of the internet.

Example Prompts

1. "Use kagi-enrich to find independent blog posts or technical write-ups about the latest developments in Rust async programming."
2. "Search the TinyGem index for non-mainstream news perspectives on recent advancements in local LLM deployment."
3. "Look for niche, non-commercial documentation on SQLite internals using the web index and limit the results to 5 items."

Tips & Limitations

• Default Index: If no index is specified, the skill defaults to web (Teclis).
• Cost: Be aware that every non-zero result costs money. Keep your queries specific to avoid unnecessary API spend.
• Timeouts: For complex, slow-loading sources, you can increase the --timeout flag, though 15 seconds is usually sufficient for standard queries.
• Performance: Because this uses a Go-compiled binary, it is extremely efficient and lightweight, making it suitable for high-frequency agent tasks.