Token Meter

Open-source MCP server for LLM spend tracking, cost estimation, and provider comparison.

tokenmeter.sh | Docs | Dashboard

What it does

Token Meter tracks your LLM API costs across 10+ providers. Compare model pricing, estimate costs before you spend, set budget alerts, and route requests to the cheapest provider — all from your AI editor.

Works as a local MCP server (open source), remote MCP server (hosted), API gateway, and dashboard.

Quick Start

Local MCP Server (no account needed)

npx @yawlabs/tokenmeter-mcp

This runs the MCP server locally with in-memory storage. All Community tier tools are available immediately — no API key, no sign-up.

Add to Claude Code

claude mcp add tokenmeter -- npx @yawlabs/tokenmeter-mcp

Add to Cursor / VS Code

Add to your MCP client config:

{
  "mcpServers": {
    "tokenmeter": {
      "command": "npx",
      "args": ["@yawlabs/tokenmeter-mcp"]
    }
  }
}

Remote MCP (hosted, no install)

Sign up at tokenmeter.sh and add to your MCP config:

{
  "mcpServers": {
    "tokenmeter": {
      "url": "https://mcp.tokenmeter.sh",
      "headers": {
        "Authorization": "Bearer tm_your_api_key"
      }
    }
  }
}

API Gateway

Swap your provider base URL for smart routing and automatic failover:

OPENAI_BASE_URL=https://gateway.tokenmeter.sh/v1

Tools

Community (Free, Open Source)

Available locally with no account:

| Tool | Description | |------|-------------| | get_spend_summary | Total spend for today, this week, and this month by provider | | get_session_cost | Cost of the current conversation/session | | get_model_pricing | Per-token pricing for any model or provider | | get_cost_estimate | Estimate cost for a workload across multiple models | | compare_models | Side-by-side cost comparison for a specific workload | | get_budget_status | Check spend against configured budget limits | | get_providers | List supported LLM providers |

Pro ($12/mo)

Requires a tokenmeter.sh account:

| Tool | Description | |------|-------------| | get_spend_breakdown | Detailed spend by provider, model, and project | | get_cost_trend | Daily spend over time with trend analysis | | get_top_models | Rank models by spend or token volume | | get_anomalies | Flag unusual spend spikes | | set_budget_alert | Create spend threshold notifications | | set_budget_cap | Set hard spending caps that block requests | | delete_budget_alert | Delete a budget alert by name or ID | | tag_session | Label sessions by project for cost attribution | | export_report | Generate spend reports (JSON/CSV) | | get_rate_limits | Rate limit status per provider | | get_routing_status | Provider health and routing status | | set_routing_rule | Configure smart routing (cheapest, fastest) | | set_fallback_chain | Define provider failover order | | set_project_quota | Set per-project spend caps | | get_latency_report | P50/P95/P99 latency per provider | | set_model_alias | Map friendly names to provider/model pairs | | get_cache_stats | Semantic cache hit rate and cost savings | | set_cache_config | Enable/disable caching, adjust TTL |

Team ($39/mo per seat)

| Tool | Description | |------|-------------| | get_team_spend | Aggregated spend across all team members | | set_team_budget | Organization-level budget caps |

Supported Providers

Anthropic (Claude), OpenAI (GPT, o-series), Google (Gemini), Groq, DeepSeek, Mistral, Cohere, Ollama, Azure OpenAI, AWS Bedrock

Development

# Install dependencies
npm install

# Run locally (stdio MCP server)
npm run dev

# Run tests
npm test

# Build
npm run build

Architecture

src/
  index.ts          # Local MCP entry point (stdio)
  mcp/server.ts     # Tool definitions and handlers
  db/memory.ts      # In-memory data store (local)
  db/postgres.ts    # PostgreSQL store (cloud)
  pricing/          # Model pricing engine + YAML data
  tools/            # Tool handler implementations
  gateway/          # API gateway + smart routing
  api/              # REST API server (cloud)
  billing/          # LemonSqueezy integration
  auth/             # Auth + team management

The local MCP server (src/index.ts) uses MemoryStore and requires zero external infrastructure — no database, no Redis, no API keys. Data is ephemeral (in-process memory).

The cloud service adds persistent storage (PostgreSQL), caching (Valkey), authentication, billing, and the web dashboard.

Part of the Yaw Labs ecosystem

• Yaw Terminal — Multi-provider AI terminal
• mcp.hosting — Deploy MCP servers in one click
• ctxlint — Lint context to reduce token waste
• tailscale-mcp — MCP servers over private tailnets
• Token Limit News — Weekly AI tooling newsletter

License

Copyright (c) 2026 Yaw Labs LLC. All rights reserved.