Access Kagi search and summarizer via session tokens — completely free. Works with Claude Desktop, Cursor, Windsurf, Cline, Hermes and any MCP-compatible AI coding assistant/agent.

⭐ Star History

⚠️ EndUser Warning: This project uses Kagi session tokens to access search and summarizer features without the official paid API. This may violate Kagi's Terms of Service. Potential consequences include account suspension or permanent ban. Use at your own risk. The authors assume no liability for any account actions taken by Kagi.

⚠️ Security Warning: Your Kagi session token provides full account access. Treat it like a password. Never share it or commit it to version control.

🌟 Features

• Session-based access: Uses Kagi session tokens instead of the official API ($25/1000 queries) — completely free
• Multi-token pool: Configure multiple tokens for higher throughput with round-robin rotation
• Per-token rate limiting: 5 req/s per token with token bucket algorithm
• API-compatible output: Returns results in the same format as the official kagimcp
• Two tools: kagi_search_fetch (web search) and kagi_summarizer (URL summarization)
• Firefox UA spoofing: Mimics a real Firefox browser for session-based access
• Auto-disable expired tokens: Expired tokens are automatically detected and skipped
• Search operators: Supports Kagi operators (site:, lang:, filetype:, etc.)

Compatible Platforms

Works with any MCP-compatible client:

• Claude Desktop
• Cursor
• Windsurf
• Cline
• Hermes Agent
• Any MCP client supporting stdio or HTTP transport

Quick Install (For AI Agents)

Just paste this prompt to your AI agent:

Install the Kagi Session2API MCP server from https://pypi.org/project/kagi-session2api-mcp/.
Then add it to my MCP:
`uvx kagi-session2api-mcp`
set env KAGI_SESSION_TOKEN to my token (ask me if you don't have it).
If I have multiple tokens, use KAGI_SESSION_TOKENS (comma-separated) instead,
or create a config file at ~/.config/kagi-session2api-mcp/config.toml with:
[kagi]
session_tokens = ["TOKEN_1", "TOKEN_2"]
summarizer_engine = "cecil"
[client]
timeout = 30
max_retries = 2
and set env KAGI_SESSION_CONFIG to that path.

Install Manually

pip install kagi-session2api-mcp

Or with uvx:

uvx kagi-session2api-mcp

Configuration

Option 1: Environment Variable (Single Token)

{
  "mcpServers": {
    "kagi-session": {
      "command": "uvx",
      "args": ["kagi-session2api-mcp"],
      "env": {
        "KAGI_SESSION_TOKEN": "YOUR_SESSION_TOKEN_HERE"
      }
    }
  }
}

Option 2: Environment Variable (Multiple Tokens)

{
  "mcpServers": {
    "kagi-session": {
      "command": "uvx",
      "args": ["kagi-session2api-mcp"],
      "env": {
        "KAGI_SESSION_TOKENS": "TOKEN_1,TOKEN_

... [View full README on GitHub](https://github.com/KSroido/Kagi-Session2API-MCP#readme)