Skip to main content
There are two ways to integrate Ceramic search into OpenClaw: a plugin and a skill. The plugin is recommended — it registers ceramic_search as a native agent tool, handles rewriting natural language queries into keyword queries internally via an LLM call, and runs searches in parallel with deduplication. This produces more consistent and higher-quality results than the prompt-based skill approach. OpenClaw agents also generally remember to call Ceramic search more with the plugin, as skills are merely system prompt-level suggestions that agents aren’t required to follow. The skill is simpler in that the instructions for query rewriting are included in the skill description, and there is no additional LLM call. The skill may also be easier to customize.

​
Setup

​
1. API key

Obtain and set your Ceramic API key:

Get API Key

Create a Ceramic account for free to get an API key.
To persist it across sessions, add the line above to your ~/.zshrc, ~/.bashrc, or equivalent. If using the plugin, you can also do:

​
2. Installation

To install the plugin, run:
To install the skill instead, run:

​
3. Allow the plugin and expose the tool

If using the plugin, do:

​
4. Modify TOOLS.md

Add the following to ~/.openclaw/workspace/TOOLS.md (create the file if it doesn’t exist) to ensure Ceramic search gets used:
Replace “ceramic_search tool” with “ceramic-search skill” if using the skill instead of the plugin.

​
5. Disable competing web search tools

This prevents the agent from using the built-in model search and forces it to use Ceramic search instead.

​
6. Restart the gateway

​
7. Test it

A successful run will show Ceramic search results via citations in the agent response.

ClawHub Plugin

View plugin

ClawHub Skill

View skill

GitHub

View source code