Thinkgate MCP server

ThinkGate

Automatic reasoning mode selection for Claude agents.

The problem

You built an AI agent. It handles everything — status checks, quick lookups, complex architecture questions, deep debugging sessions. But under the hood it runs every single message through the same model with the same thinking settings.

That means you're burning extended thinking tokens on "what time is it in Tokyo?" and getting shallow answers on "help me design the entire auth system."

You could manually tag requests — ULTRATHINK: before the hard ones. But you forget. Your users definitely won't do it. And if you're building agents for other people, you can't train every end user to manage thinking modes.

ThinkGate fixes this at the infrastructure layer. It sits between the incoming message and your model call, classifies the complexity in ~200ms, and returns exactly which model and thinking depth to use. Automatically. Every time.

Who this is for

• Agent builders running Claude on a mix of simple and complex tasks who are tired of one-size-fits-all model settings
• Teams running 24/7 agents (WhatsApp bots, Slack assistants, Telegram agents) where message complexity varies wildly and cost/latency actually matters
• Anyone who's ever typed ULTRATHINK manually and thought: this should just happen on its own

How it works

Incoming message
      ↓
  Haiku call (~200ms, ~$0.0001)
  "How complex is this?"
      ↓
  fast → no extended thinking
  think → medium effort
  ultrathink → max effort
      ↓
  Claude runs with the right settings

A cheap, fast Haiku call reads your prompt and decides which tier it needs. Then your main Claude call runs with the right effort level. You pay almost nothing for the classification, and save real money (and latency) on the 60%+ of messages that don't need extended reasoning.

The classifier is the IP here — not which model runs it. Three tiers. A system prompt trained on the boundary between "this needs thinking" and "this doesn't." Works out of the box.

Tiers

Profiles (model mapping)

Set THINKGATE_PROFILE or pass { profile } to classifyPrompt:

Per-tier overrides: THINKGATEFASTMODEL, THINKGATETHINKMODEL, THINKGATEULTRAMODEL.

Cost note: the MCP tool is advisory. Hosts must actually switch models. Pi does this via ~/.pi/agent/extensions/thinkgate-router.ts + thinkgate.json.

Rule mode (no API key) is free and preferred for auto-routing. Long checklist pastes no longer auto-upgrade to THINK.

Use as an MCP tool (Claude Desktop / Claude Code)

Add to ~/.claude/settings.json (Claude Code) or ~/Library/Application Support/Claude/claudedesktopconfig.json (Claude Desktop):

{
  "mcpServers": {
    "thinkgate": {
      "command": "npx",
      "args": ["-y", "mcp-thinkgate"],
      "env": {
        "ANTHROPIC_API_KEY": "your-api-key-here"
      }
    }
  }
}

Restart Claude. Now you can ask it to classify before it answers:

Shortened. The full README is on GitHub.