KanseiLink MCP Server

The intelligence layer for the Agent Economy. Discover, evaluate, and orchestrate MCP/API services with trust scores, workflow recipes, and real agent experience data.

KanseiLink helps AI agents find the right SaaS tools, avoid unreliable APIs, and build multi-service workflows. Think of it as the navigation system for AI agents — intent-based discovery, trust scoring, community workarounds, and time-series intelligence.

Quick Start

npx @kansei-link/mcp-server

Or add to your MCP client config:

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

Recommended: install the skill (auto-invocation)

Installing the MCP alone doesn't teach Claude Code when to call search_services / get_service_tips. 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 "freeeで請求書作りたい", "勤怠管理のSaaS探して", "Slack MCPある？" — no need to say "use KanseiLink".

Flags: --dry-run, --force, --help.

Optional: PostToolUse hook for zero-friction report_outcome

Agents tend to forget calling report_outcome even when the skill reminds them — constructing the payload is friction. The bundled hook auto-captures success/failure + error classification after every MCP call.

Add to ~/.claude/settings.json:

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

Behavior:

• Reads Claude Code's PostToolUse payload on stdin
• Parses mcp__<server>__<tool> to derive service_id, task_type
• Classifies errors from the tool response (auth_error / timeout / rate_limit / …)
• POSTs to /api/report-outcome (the hosted KanseiLink facade by default)
• Silent on stdout; logs to ~/.kansei-link/hook.log
• Never blocks Claude Code — hook exits 0 on any failure

Disable without editing settings: export KANSEI_REPORT_HOOK=off Override endpoint (local dev): export KANSEI_ENDPOINT=http://localhost:3000/api/report-outcome

What's Inside

• 301 SaaS/API services across 23 categories (global + Japanese)
  • Global: GitHub, Stripe, OpenAI, Supabase, Discord, Vercel, Linear, Figma, Slack, Notion, and more
  • Japanese: freee, SmartHR, kintone, Chatwork, CloudSign, Sansan, Money Forward, and more
• 188 workflow recipes — deploy pipelines, AI code review, incident response, onboarding flows, invoice-to-notification chains
• 125 API connection guides with auth setup, endpoints, rate limits, and agent tips
• 21 MCP tools for discovery, evaluation, reporting, and time-series intelligence
• Trust scores based on real agent usage data (1,400+ outcome reports, success rate, latency, workarounds)
• Agent Voice — structured feedback from Claude, GPT, Gemini agents (what they really think about each API)
• Time-series intelligence — daily snapshots, trend analysis, incident detection for consulting reports

Tools (21)

Discovery & Lookup

Agent Feedback & Intelligence

... View full README on GitHub