Keenable Web Search

Live web search and clean page extraction via the Keenable MCP server. Use when you need to: (1) search the web for current, citable information beyond the model's training data, (2) find news, research, docs, or specific pages by describing them in natural language, (3) restrict results to a site or a date range, (4) fetch a known URL as clean markdown. Works with no API key. NOT for: local file operations or anything that does not involve web search or page retrieval.

Keenable — Web Search & Page Extraction

Keenable is a hosted web-search engine exposed over MCP. Describe the page you want in natural language and it returns ranked results with titles, URLs, and snippets, or fetches a known URL as clean, LLM-ready markdown.

MCP server: https://api.keenable.ai/mcp (Streamable HTTP) Free tier: works keyless — a rate-limited public endpoint, no signup API key: set the X-API-Key header to raise rate limits (and to enable the low-latency realtime search mode on eligible accounts) — keenable.ai Site: keenable.ai

Setup

Add the MCP server to your agent config:

bash

# OpenClaw — keyless
openclaw mcp add keenable --url "https://api.keenable.ai/mcp"

Or in any MCP config JSON (Streamable HTTP transport):

json

{
  "mcpServers": {
    "keenable": {
      "url": "https://api.keenable.ai/mcp"
    }
  }
}

To raise rate limits, send your key as an X-API-Key header:

json

{
  "mcpServers": {
    "keenable": {
      "url": "https://api.keenable.ai/mcp",
      "headers": { "X-API-Key": "YOUR_KEENABLE_API_KEY" }
    }
  }
}

---

Tool Reference

Tool	What it does
search_web_pages	Web search — natural-language query in, ranked results (title, URL, snippet) out
fetch_page_content	Fetch a known URL and return the page as clean markdown

---

search_web_pages

Describe the page you want in natural language, not just keywords — Keenable ranks on meaning. Snippets are starting points; open the URL with fetch_page_content before relying on a claim, and cite the source.

Parameters:

Parameter	Type	What it does
query	string, required	Semantically rich description of the ideal page (e.g. "official changelog for the EU AI Act 2025 amendments"), not bare keywords
mode	pro | realtime	pro (default) for higher-quality results; realtime for the lowest latency (good for voice / latency-sensitive turns). realtime requires an enabled account
site	string	Restrict results to one site, e.g. arxiv.org
published_after / published_before	string	YYYY-MM-DD bounds on publish date — use for time-sensitive topics
acquired_after / acquired_before	string	YYYY-MM-DD bounds on when the page was indexed

text

search_web_pages {
  "query": "blog posts comparing vector databases for recommendation systems",
  "site": "arxiv.org"
}

text

search_web_pages {
  "query": "latest OpenAI model announcements",
  "published_after": "2026-01-01",
  "mode": "realtime"
}

Search modes

• pro — default. Higher-quality ranking. Best for research and accuracy.
• realtime — fastest response, ideal for latency-sensitive turns (e.g. voice agents). Requires an account with realtime enabled; keyless and standard accounts should use pro.

---

fetch_page_content

Fetch a known URL and get the page back as clean markdown — no scraping or HTML parsing on your side. Use it to read a result returned by search_web_pages before citing it.

Parameters:

Parameter	Type	What it does
url	string, required	The URL to fetch, e.g. https://example.com
max_chars	int	Cap the returned content length. Longer pages are truncated. Defaults to 50000

text

fetch_page_content {
  "url": "https://example.com/article",
  "max_chars": 20000
}

---

Typical flow

1. search_web_pages with a descriptive query (add site / date filters to narrow).
2. Pick the most relevant result.
3. fetch_page_content on its URL to read the full page.
4. Answer from the fetched content and cite the source URL.