Claude

Add Keenable to Claude and Claude Code, and keep Claude’s own web search from competing with it.

MCP connector

Keenable is listed in Claude’s Connectors Directory as Keenable Web Search, so on claude.ai, Claude Desktop, Cowork and mobile it takes a few clicks and no config file. Claude Code adds the same server with one command. Either way Claude also has a web search of its own, and with both available it picks between them from one lookup to the next. Each section below covers adding Keenable, then turning Claude’s search off, and how to turn it back on.

Claude (web, Desktop, mobile)

Add the connector

1. Find it in the directory

   Open Customize → Connectors, click +, and search for Keenable Web Search. From a chat you get to the same place through + → Connectors → Manage connectors.

2. Connect

   Click Connect. There is no sign-in and nothing to paste.

3. Turn it on in the chat

   Click + in the lower left of the message box, hover Connectors, and toggle Keenable Web Search on. Connectors are switched on per conversation, so if Claude answers without ever calling Keenable, this is the step that was skipped.

The listed connector runs on the shared public tier at lower rate limits. It works in Cowork tasks as well as in chats, so a longer task can search and read pages while it works with your files and other connectors.

Use your own account

To run on your own Keenable account, with its rate limits and billing, add Keenable by URL instead of from the directory and sign in:

1. Customize → Connectors → + Add → Add custom connector.

2. Name it Keenable, enter this URL, and click Add:

   https://api.keenable.ai/mcp-oauth
   

3. Click Connect, sign in to Keenable, and allow access.

Searches then count against your current organization in Keenable. On Team and Enterprise plans an Owner adds the connector once under Organization settings → Connectors, and each member connects it and signs in to their own account.

You can pass an API key instead of signing in: add https://api.keenable.ai/mcp, choose No sign-in under Authentication, and under Request headers pick x-api-key with your key (keen_<your_key>, from the console). Anthropic is rolling request headers out in beta to a limited set of organizations, and a connector’s authentication can’t be edited after it is added, so to change the key, remove the connector and add it again.

Turn off web search

On a Team or Enterprise plan, an Owner turns it off once for everyone: open Organization settings → Capabilities and switch Web search off under Data sources. Members then see it as set by the admin, and Claude has Keenable as its only way to search in every chat. Switch it back on in the same place.

On a Pro or Max plan, it depends on which version of the Claude interface your account has.

If your message box offers Chat and Cowork options, web search is a switch in each chat. Click + in the lower left of the message box, then click Web search so its checkmark disappears. Click it again to turn it back on.

If it doesn’t (the newer interface, rolling out to Pro and Max first), there is no web search switch and Claude decides when to search on its own. You can still tell it which tool to use: open Settings → General, and add this to Instructions for Claude:

For web searches and reading web pages, use the Keenable Web Search connector (search_web_pages and fetch_page_content) instead of the built-in web search.

Delete the line to go back. If Claude still doesn’t reach for Keenable, open + → Connectors → Tool access in the chat and choose Always available, which loads every connector’s tools at the start of the conversation.

Claude Code

Add the server

claude mcp add --transport http --scope user keenable https://api.keenable.ai/mcp \
  --header "X-API-Key: keen_<your_key>"

--scope user makes it available in every project. Omit the --header flag to run on the public tier.

Turn off the built-in search

Claude Code ships its own WebSearch and WebFetch tools. Deny them in ~/.claude/settings.json:

{
  "permissions": {
    "deny": ["WebSearch", "WebFetch"]
  }
}

A rule that names a whole tool removes it from Claude’s context, so Keenable’s search_web_pages and fetch_page_content are the only search and fetch tools it has left. If the file already has a permissions block, add the two names to its deny list rather than adding a second block. Drop "WebFetch" if you want to keep Claude’s own page reader.

To turn the built-in search back on, delete the entries. To try it for a single session without editing the file:

claude --disallowedTools WebSearch WebFetch

Verify

Run /mcp to confirm keenable is connected with both tools listed, then ask something that needs the live web. Claude should call mcp__keenable__search_web_pages without being told to.

What to ask

Keenable gives Claude two tools, and Claude combines them across a longer task.

Search

Ranked results with an excerpt of each page’s text, narrowed to one site or a date range when the question calls for it.

Read

Any URL as clean markdown: documentation, changelogs, issues, papers, pricing pages.

Research

Several searches, the pages worth reading, and an answer that cites them.

Ask for the result you want, and say which sources count:

Find the release notes for every version of zod between the one in package.json
and the latest, list the breaking changes, and update our schemas to match.

Before you change anything, look for reports of this error on github.com from
the last 30 days and tell me whether there is a known fix.

Check each price in this comparison table against the vendor's own pricing page
today, fix what is out of date, and link the page next to every number.

The first two suit Claude Code, where Claude can act on what it finds in your repository. The third works in a chat or a Cowork task with the table attached.

Keenable SELECT

Keenable SELECT is listed in the same directory as a second connector. It is a separate server that answers a question with one SQL query over live web results and returns a table; see SELECT.

→ Full MCP server reference