How many sources does the research agent search?

The default SOUL.md instruction in this module tells the agent to search at least 3 different sources. You can raise this to 5 or 10 by updating the instruction. For deeper research, use the `--thinking high` flag — it makes the agent reason more carefully before deciding which sources to trust.

Does the research agent save its findings automatically?

Yes, when you include the notes skill and instruct the agent to save summaries in SOUL.md. The agent saves findings as a Markdown file named after the topic. You can retrieve past research by asking the agent to 'find my notes about X' or by opening the file directly in your workspace.

Can I ask the agent to research something while I'm away?

Yes. Use the CLI to send a message without waiting for a response: `openclaw agent --message 'Research X and save a summary to my notes' --async`. The agent runs the task in the background and stores the result. You can check the notes later by asking 'what did you find about X?'

Project: Research Companion

Build an AI research agent that searches the web, reads articles, summarizes findings, and saves organized notes — all from a single message.

0 of 10 modules complete0%

3 min read

What you will build

A research agent that can take a topic, search multiple sources, summarize findings, and save organized notes — all from a single message.

Install research skills

openclaw skills install web-search
openclaw skills install web-reader
openclaw skills install notes
openclaw skills install summarizer

Configure research behavior

Add to your SOUL.md:

## Research mode
When I ask you to research a topic:
1. Search at least 3 different sources
2. Read the most relevant pages in full
3. Summarize key findings in bullet points
4. Include direct quotes for important claims
5. List all sources at the end with URLs
6. Save the summary to my notes with the topic as the title

Test a research task

openclaw agent --message "Research the current state of solar panel efficiency. What are the latest breakthroughs?" --thinking high

--thinking high tells the agent to reason more carefully before responding — it takes a few extra seconds but produces better results for complex tasks like research. For simple questions you don't need it.

The agent should search the web, read several articles, and give you a structured summary with sources.

Organize findings

Ask the agent to build on previous research:

Based on my saved notes about solar panels, what are the 3 most
promising technologies for residential use?

The agent uses its memory and saved notes to give a deeper, more informed answer.

Advanced: Sub-agents

For complex research, you can configure sub-agents that specialize in different aspects. See Module 6 for sub-agent setup.

Tips

Use --thinking high for better source evaluation
Ask for comparisons: "Compare X vs Y based on my research"
Export notes: openclaw notes export --format markdown

Finished this module?

Tracks your progress across all 10 modules

FAQ

How many sources does the research agent search?: The default SOUL.md instruction in this module tells the agent to search at least 3 different sources. You can raise this to 5 or 10 by updating the instruction. For deeper research, use the `--thinking high` flag — it makes the agent reason more carefully before deciding which sources to trust.
Does the research agent save its findings automatically?: Yes, when you include the notes skill and instruct the agent to save summaries in SOUL.md. The agent saves findings as a Markdown file named after the topic. You can retrieve past research by asking the agent to 'find my notes about X' or by opening the file directly in your workspace.
Can I ask the agent to research something while I'm away?: Yes. Use the CLI to send a message without waiting for a response: `openclaw agent --message 'Research X and save a summary to my notes' --async`. The agent runs the task in the background and stores the result. You can check the notes later by asking 'what did you find about X?'