Keenable Web Search

Keenable Web Search is, according to the provider, an independent remotely operated Model Context Protocol offering for web search and page content. The official product page at https://keenable.ai describes Keenable as independent search infrastructure with its own crawler, index, and ranking. The official documentation at https://docs.keenable.ai/mcp-server describes the MCP service and names the stable endpoint https://api.keenable.ai/mcp. Smithery lists keenable/web-search as a verified, remote, and deployed server. Smithery also exposes the deployment endpoint https://web-search-keenable.run.tools, which forwards to the Keenable endpoint with a Smithery title parameter. The product page, provider documentation, Smithery source, and remote identity therefore establish an independently catalogable durable product rather than a temporary demo. No public provider repository is published, so this entry stores neither an invented repository link nor a star count.

Purpose and documented scope

According to the provider, the server exposes two read-oriented tools. search_web_pages returns relevant results with URLs, titles, descriptions, and ranking. The documentation describes filters for individual sites, publication or acquisition periods, and point-in-time search. An agent can therefore limit research to a domain, a time window, or a deliberate freshness requirement. fetch_page_content retrieves a URL and extracts its content as Markdown. By default, only URLs from the index are supported; the documented live option can access a source directly and should therefore be treated as broader access. The service is not a general browser, does not guarantee completeness, and does not replace primary-source review.

The product page places the MCP capabilities within Keenable search infrastructure; the integrations documentation names the same endpoint for MCP-compatible clients and links back to Keenable documentation. Smithery records readOnlyHint and destructiveHint false for both tools. Those are useful technical signals, but they are not a complete security guarantee. Search results, snippets, fetched pages, redirects, and metadata remain untrusted input. They can be stale, incomplete, copyrighted, misleading, or capable of prompt injection. Source quality must be assessed using author, publisher, date, original URL, methodology, and independent confirmation. For scientific, legal, medical, financial, or security-sensitive questions, the responsible primary source remains authoritative.

Authorization, privacy, and data sharing

The provider documentation describes public access without a visible key and an optional API key for other access conditions. The absence of visible authentication is not a promise of unlimited use, unchanged rate limits, or permanent availability. Operators should review current documentation, terms, and reachable hostnames, enforce TLS, and allow only the expected endpoint. This catalog stores no tokens, API keys, secrets, or credentials. Search queries can reveal sensitive intent, internal project names, personal references, or confidential URLs. Before use, review data minimization, redaction, roles, retention, deletion, storage region, log access, and subprocessors.

A remote request is processed at the Keenable endpoint. The MCP client, model provider, logs, telemetry, or a Smithery proxy may also receive data. Sharing with those parties must be traceable in the relevant configuration and privacy assessment. A local client does not automatically make processing local. URLs containing credentials, personal search text, and internal document paths should not be sent as queries.

Access restrictions, write access, and security

The documented tools are read-oriented and non-destructive according to Smithery. That does not prevent write access outside the server: an agent can transfer results into files, notes, databases, tickets, or publications. Those destinations require separate permissions, least privilege, allowlisting, and human approval. The live option in fetch_page_content expands access from indexed pages to direct sources and should be used only with deliberately approved domains. Redirects, oversized responses, malicious links, and untrusted Markdown require timeouts, size limits, content inspection, and safe output isolation.

The E-E-A-T basis of this entry is, according to the provider, the product page and Keenable documentation, supplemented by Smithery metadata for verified deployment, endpoint, and tool hints. These sources support operator identity, product boundary, and documented scope, not the truth of every search result. Before publication or consequential use, open the original sources, check freshness, attribute quotations correctly, and handle conflicting evidence transparently.

An MCP-compatible client with HTTPS access to the official endpoint. Optional access tiers may have current provider requirements.

Configure the official remote endpoint https://api.keenable.ai/mcp in the MCP client according to current Keenable documentation. Smithery users may use the deployed link, but should review the proxy, host, and data sharing.

https://api.keenable.ai/mcp

The documentation describes public access and an optional API key for other access conditions. Verify current authentication, limits, and terms with the provider; never store secrets in the catalog.

Search and retrieval tools are read-oriented and non-destructive according to Smithery. Domain allowlisting, TLS, and separate approval for external write destinations remain necessary.

Queries and responses reach the remote endpoint; the client, model provider, logs, telemetry, or Smithery may process data as well. Minimize search text and URLs and review retention.

Prompt injection, stale or incorrect results, sensitive inferences, copyrighted content, redirects, and unintended external writes. Mitigate with source review, limits, allowlisting, and approvals.