Kagi Search
MCP ServerFree** - Search the web using Kagi's search API
Capabilities5 decomposed
web-search-via-kagi-api
Medium confidenceExecutes web searches through Kagi's proprietary search API, returning ranked results with titles, snippets, and URLs. The MCP server translates search queries into Kagi API calls, handling authentication via API keys and formatting responses into structured JSON that Claude or other MCP clients can consume. Implements request batching and result pagination to support both single queries and multi-step search workflows.
Implements Kagi search as an MCP server, enabling Claude and other MCP clients to invoke web search as a native tool without custom API wrappers. Uses Kagi's privacy-focused search index (no user tracking) and integrates directly into the MCP protocol's function-calling mechanism, allowing declarative search composition within agent workflows.
Offers privacy-first search integration for MCP clients (unlike Google/Bing APIs which track users), and provides direct Claude compatibility without requiring custom tool definitions or API orchestration code.
mcp-protocol-search-tool-binding
Medium confidenceWraps Kagi search functionality as an MCP-compliant tool that Claude and other MCP clients can discover and invoke through the Model Context Protocol. The server exposes search as a callable function with schema validation, parameter marshalling, and response serialization following MCP's tool definition standard. Handles tool discovery, schema advertisement, and request/response lifecycle management within the MCP message protocol.
Implements full MCP tool protocol compliance, including schema advertisement, parameter validation, and error handling within the MCP message lifecycle. Unlike generic API wrappers, this exposes search as a first-class MCP tool that Claude can discover and invoke with natural language, enabling seamless integration into agent reasoning loops.
Provides native MCP integration (vs. custom tool definitions), enabling Claude to automatically invoke search without explicit prompt engineering, and allows tool discovery and composition within the MCP ecosystem.
kagi-api-authentication-and-rate-limiting
Medium confidenceManages Kagi API authentication by storing and validating API keys, implementing request signing if required by Kagi's API, and enforcing rate limits to prevent quota exhaustion. The MCP server handles credential injection into outbound requests, token refresh if applicable, and graceful degradation when rate limits are exceeded. Implements exponential backoff for retries and tracks quota usage across multiple concurrent search requests.
Implements MCP-native credential handling where API keys are managed by the MCP server process, not by the client, ensuring keys are never exposed to Claude or other MCP clients. Uses environment-based configuration for secure key storage and implements client-side rate limiting with exponential backoff to prevent quota exhaustion.
Separates credential management from client logic (vs. embedding keys in prompts or client code), and provides rate-limit protection without requiring manual quota tracking by the application.
search-result-parsing-and-formatting
Medium confidenceParses Kagi API responses and transforms raw search results into a standardized JSON format suitable for LLM consumption. Extracts relevant fields (title, snippet, URL, domain), sanitizes HTML entities and special characters, truncates long snippets to fit context windows, and structures results as an array with consistent schema. Handles edge cases like missing fields, malformed responses, and encoding issues.
Implements LLM-aware result formatting that prioritizes snippet clarity and token efficiency, including automatic truncation and domain extraction. Unlike generic API response passthrough, this normalizes Kagi's response schema into a format optimized for Claude's context window and reasoning capabilities.
Provides LLM-optimized formatting (vs. raw API responses), with automatic snippet truncation and domain extraction, reducing the need for post-processing in agent code.
multi-query-search-composition
Medium confidenceEnables agents to compose multiple sequential or parallel search queries, where results from one query can inform subsequent queries. The MCP server maintains request context across multiple tool invocations, allowing agents to refine searches based on intermediate results. Implements query deduplication to avoid redundant API calls and result caching within a single agent session to reduce API usage and latency.
Implements session-scoped result caching and query deduplication within the MCP server, allowing agents to perform multi-step research without redundant API calls. Unlike stateless search APIs, this maintains context across multiple tool invocations, enabling intelligent query refinement and result synthesis.
Provides built-in caching and deduplication (vs. agents managing their own state), reducing API calls and latency for multi-step research workflows.
Capabilities are decomposed by AI analysis. Each maps to specific user intents and improves with match feedback.
Related Artifactssharing capabilities
Artifacts that share capabilities with Kagi Search, ranked by overlap. Discovered automatically through the match graph.
Kagi Search
Premium ad-free search engine with AI summarization.
Kagi
** - Kagi search API integration
Kagi
Premium ad-free search — AI summarization, custom ranking, privacy-respecting, FastGPT.
Google PSE/CSE
** - A Model Context Protocol (MCP) server providing access to Google Programmable Search Engine (PSE) and Custom Search Engine (CSE).
Search1API
** - One API for Search, Crawling, and Sitemaps
kisti-mcp
MCP server: kisti-mcp
Best For
- ✓AI agent builders using Claude with MCP protocol
- ✓Teams building search-augmented LLM applications
- ✓Developers wanting privacy-focused search without Google tracking
- ✓Solo developers prototyping research assistants or fact-checking bots
- ✓Claude Desktop users extending Claude with web search capability
- ✓MCP framework developers building tool ecosystems
- ✓Teams standardizing on MCP for LLM tool integration
- ✓Developers building agentic systems that need declarative tool composition
Known Limitations
- ⚠Requires active Kagi subscription or API credits — free tier may have rate limits
- ⚠Search results are only as current as Kagi's index refresh cycle (typically hours to days old)
- ⚠No built-in result filtering by date, domain, or content type — requires post-processing
- ⚠MCP protocol overhead adds ~100-200ms latency per search request vs direct API calls
- ⚠Limited to text-based queries — no image search or advanced query syntax beyond Kagi's native support
- ⚠Tool invocation latency depends on MCP server startup and message serialization overhead
Requirements
Input / Output
UnfragileRank
UnfragileRank is computed from adoption signals, documentation quality, ecosystem connectivity, match graph feedback, and freshness. No artifact can pay for a higher rank.
About
** - Search the web using Kagi's search API
Categories
Alternatives to Kagi Search
Search the Supabase docs for up-to-date guidance and troubleshoot errors quickly. Manage organizations, projects, databases, and Edge Functions, including migrations, SQL, logs, advisors, keys, and type generation, in one flow. Create and manage development branches to iterate safely, confirm costs
Compare →Are you the builder of Kagi Search?
Claim this artifact to get a verified badge, access match analytics, see which intents users search for, and manage your listing.
Get the weekly brief
New tools, rising stars, and what's actually worth your time. No spam.
Data Sources
Looking for something else?
Search →