Versions
Use v0.27.0 or later. In earlier releases, each Keenable result reached the agent with an empty content field.
Add the tool to a workflow
1
Open an agent
In RAGFlow, open an agent in the visual agent builder, or create one.
2
Add Keenable
Add Keenable as a tool on an Agent component, or as its own component on the canvas.
3
Fill in the form
Leave API Key empty for the keyless tier. Set the fields below, then save the agent.
The tool has two outputs.
formalized_content holds the results as text for an LLM prompt, and json holds the raw result list. Each result in formalized_content looks like this:
Check the connection
Run the request the tool sends, from the machine that hosts RAGFlow. It confirms the server can reachapi.keenable.ai over HTTPS:
results array. In the agent, ask a question that needs the web and open the run log. You’ll see the Keenable call and the pages it returned.
Troubleshooting
The error text depends on the release. v0.26.2 to v0.27.x run the Python tool, and v1.0.0 and later run the Go tool.
For another RAG platform with Keenable built in, see WeKnora.
Paste a
keen_… key in the tool’s API Key field to remove the hourly request cap. If you leave the field empty, calls run on the shared keyless tier at lower rate limits. RAGFlow doesn’t read KEENABLE_API_KEY from the environment.Related pages
Replit
Give Replit Agent web search and page fetch through the Keenable MCP server: one click, then sign in to Keenable.
n8n
Search the web and fetch pages in n8n workflows and AI Agent tools with the Keenable community node. Works without an API key.
Tines
Call Keenable web search and page fetch from Tines stories with the HTTP Request action, with no key to start.
Fetch
Fetch a web page with GET /v1/fetch and get its main text as markdown, or only the part a prompt asks for.