KanseiLink MCP Server

Local-first MCP navigator for AI agents. Find, evaluate, and compose MCP tools -- 80% fewer tokens vs trial-and-error.

Your agent wastes thousands of tokens every time it hits a new MCP service: searching docs, guessing auth flows, recovering from errors. KanseiLink eliminates that loop with a local SQLite DB of pre-evaluated services, trust scores, and step-by-step recipes.

Token savings: 89-97% measured across 7 services (avg ~16,800 tokens without → ~950 with KanseiLink).

Quick Start

npx @kansei-link/mcp-server

Works with Claude Code, Cursor, Cline, Zed, Windsurf -- any MCP client.

Add to your config (claude_desktop_config.json, .cursor/mcp.json, etc.):

{
  "mcpServers": {
    "kansei-link": {
      "command": "npx",
      "args": ["-y", "@kansei-link/mcp-server"]
    }
  }
}

Or with Claude Code CLI:

claude mcp add -s user kansei-link -- npx -y @kansei-link/mcp-server

What's Inside

	Count	Description
Services	11,000+	MCP servers and SaaS APIs across 23 categories (2,257 MCP-verified via handshake)
Recipes	200	Multi-service workflow compositions (standup, PR review, incident response, onboarding...)
API Guides	199	Auth setup, endpoints, rate limits, pitfalls, and workarounds
Trust Scores	Weekly	Based on automated health probes + real agent usage data

All data ships inside the npm package as a local SQLite DB. Zero API calls needed. No server dependency, no signup.

Why Not Just Read the Docs?

Without KanseiLink	With KanseiLink
web_search "freee API auth"	search_services({ intent: "send invoice" })
web_fetch docs landing page (SPA, mostly nav)	lookup({ service_id: "freee" })
web_fetch endpoint reference	Agent has auth flow, pitfalls, workarounds
web_fetch auth guide	in ~950 tokens
Trial-and-error on wrong params	First try succeeds
~16,800 tokens burned	89-97% saved

Claude Code: install the skill (auto-invocation)

Installing the MCP alone doesn't teach Claude Code when to call KanseiLink. The bundled skill fixes that:

npx -y @kansei-link/mcp-server kansei-link-install-skill

This copies a SKILL.md to ~/.claude/skills/kansei-link/. Claude Code auto-discovers it and fires the skill on phrases like "connect to Stripe", "Slack MCPある？", "send invoice via freee" -- no need to say "use KanseiLink".

Optional: PostToolUse hook

Auto-capture success/failure after every MCP call (agents tend to forget reporting).

Add to ~/.claude/settings.json:

{
  "hooks": {
    "PostToolUse": [{
      "matcher": "mcp__.*",
      "hooks": [{ "type": "command", "command": "npx -y @kansei-link/mcp-server kansei-link-report-hook" }]
    }]
  }
}

Disable anytime: export KANSEI_REPORT_HOOK=off

Tools (5)

v1.0 consolidates the tool surface from 25 individual tools into 5 unified tools with mode auto-detection.

Standard Flow (3 tools -- all you need)

search_services --> lookup --> (execute your API call) --> report

Tool	Modes	Description
search_services	--	Find services by intent (FTS5 + trigram + category boost)
lookup	8 modes	Get tips, detail, insights, recipes, combinations, history, feedback, voices
report	4 modes	Report outcomes, submit feedback, record events, share your voice

Admin Tools (2 additional)

Tool	Modes	Description
inspect	8 modes	Colony health: inspection queue, anomaly verification, update proposals, snapshots
analyze	4 modes	Analytics: token savings, cost audit, AEO reports and articles

Lookup Modes

Mode	Trigger	Example
tips (default)	service_id alone	lookup({ service_id: "freee" })
detail	detail: true	`lookup({ service_id: "freee"

... View full README on GitHub