Leanproxy Mcp

LeanProxy-MCP

The Local CLI Proxy That Slashes Your AI Token Bill by 90%+

---

The MCP Schema Tax is Killing Your AI Budget

Every MCP server you connect injects thousands of tokens into every LLM request — even when you never use it. This is the "Schema Tax":

flowchart LR
    IDE["Your IDE"] --> MCP["MCP Gateway"]

    subgraph MCP["MCP Gateway"]
        S1["GitHub (41 tools)"]
        S2["Garmin (55 tools)"]
        S3["Intervals (67 tools)"]
        S4["Stitch (22 tools)"]
    end

    MCP --> LLM["LLM Provider"]

    note1["~4,100 tokens"]
    note2["~5,500 tokens"]
    note3["~6,700 tokens"]
    note4["~2,200 tokens"]
    total["TOTAL: ~18,500 tokens"]

    S1 -.-> note1
    S2 -.-> note2
    S3 -.-> note3
    S4 -.-> note4
    note1 --- total
    note2 --- total
    note3 --- total
    note4 --- total
    total -.-> LLM

    style MCP fill:#ff6b6b,color:#fff
    style LLM fill:#ee5a5a,color:#fff

The result? You're burning tokens on tool definitions you'll never use in that session.

---

Enter LeanProxy: Your Token Firewall

LeanProxy sits between your IDE and MCP servers as a smart gateway. It loads tool schemas only when needed — reducing 18,500 tokens to ~110.

flowchart LR
    IDE["Your IDE"] --> Gateway["LeanProxy Gateway"]

    subgraph Gateway["LeanProxy Gateway"]
        Router["Router: invoke_tool + list_tools (~110 tokens)"]
        JIT["JIT Schema Loading"]
        Cache["Automatic Caching"]
        Firewall["Token Firewall"]
        Pool["Connection Pooling"]
    end

    Gateway -.->|"loads on demand"| GH["GitHub"]
    Gateway -.->|"loads on demand"| Garmin["Garmin"]
    Gateway -.->|"loads on demand"| Intervals["Intervals.icu"]
    Gateway -.->|"loads on demand"| Stitch["Stitch"]

    Router --> Firewall --> Pool --> GH
    Router --> Firewall --> Pool --> Garmin
    Router --> Firewall --> Pool --> Intervals
    Router --> Firewall --> Pool --> Stitch

    Router --> JIT
    JIT --> Cache

    style Gateway fill:#1a1a2e,color:#fff
    style Router fill:#00ADD8,color:#fff
    style JIT fill:#00ADD8,color:#fff
    style Cache fill:#00ADD8,color:#fff
    style Firewall fill:#00ADD8,color:#fff
    style Pool fill:#00ADD8,color:#fff

---

Real Results, Real Savings

90%+ Token Reduction in Production Sessions

| Session Type | Native MCP | LeanProxy | Savings | |:-------------|:-----------|:---------|:--------| | Morning Sport (2 servers, 4 prompts) | ~21,000 | ~2,000 | 90% | | Dev Workflow (2 servers, 5 prompts) | ~10,600 | ~2,400 | 77% | | Full Day (4 servers, 7 prompts) | ~49,600 | ~3,500 | 93% |

The Math Doesn't Lie

Native MCP + 100% cache hit still costs you at 0.25x (cache read isn't free!):

| Configuration | Native MCP (cached) | LeanProxy | You Save | |:--------------|:-------------------|:---------|:--------| | 1 server (41 tools) | 1,025 tokens | 27 tokens | 97.3% | | 2 servers (53 tools) | 1,325 tokens | 27 tokens | 97.9% | | 4 servers (163 tools) | 4,075 tokens | 27 tokens | 99.3% |

---

Key Features

| Feature | Benefit | |:--------|:--------| | 🛡️ Token Firewall | Redacts secrets, API keys, and PII before they reach LLM providers | | ⚡ JIT Schema Loading | Tool schemas load only when actually called — not on every request | | 🔄 Connection Pooling | HTTP MCP clients reuse connections with circuit breakers | | 📦 Multi-Transport | Supports stdio, HTTP, and SSE transport protocols | | 👥 Multi-Team Namespaces | Hierarchical organization for enterprise teams | | 💰 Cost Attribution | Track token savings per server with detailed reports | | 🧪 Dry-Run Mode | Simulate and preview savings without live execution | | 🔧 Shadow Manifesting | Merges global and project-local MCP configurations |

---

Quick Start

One-Line Install

# macOS/Linux via Homebrew
brew tap mmornati/leanproxy-mcp https://github.com/mmornati/leanproxy-mcp
brew install leanproxy-mcp

# ...or download binary for your platform
curl -fsSL https://github.com/mmornati/leanproxy-mcp/releases/latest/download/leanproxy-mcp.tar.gz | tar xz

Configure Your IDE

Add LeanProxy as an MCP server in your opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "leanproxy": {
      "type": "local",
      "command": ["leanproxy-mcp", "server", "run", "--stdio"],
      "enabled": true
    }
  }
}

Run It

# Start with your MCP servers
leanproxy-mcp server run --stdio

# Preview savings without executing
leanproxy-mcp server run --dry-run --stdio

# Generate a detailed savings report
leanproxy-mcp report --output report.md

---

Architecture

flowchart TB
    subgraph IDE["Your IDE"]
        Client[(MCP Client)]
    end

    subgraph Gateway["LeanProxy Gateway"]
        Router["Router(~110 tokens)"]
        JIT["JIT Schema Cache"]
        Firewall["Token Firewall(Secret Redaction)"]
        Pool["Connection Pool& Circuit Breaker"]
        Federation["Federation(Multi-org)"]
    end

    subgraph Servers["MCP Servers"]
        GH["GitHub(stdio)"]
        Garmin["Garmin(HTTP)"]
        Intervals["Intervals.icu(SSE)"]
        Stitch["Stitch(HTTP)"]
    end

    Client  Router
    Router  JIT
    Router  Firewall
    Firewall  Pool
    Pool  Federation

    Pool --- GH
    Pool --- Garmin
    Pool --- Intervals
    Pool --- Stitch

    JIT -.->|"loads on call"| GH
    JIT -.->|"loads on call"| Garmin
    JIT -.->|"loads on call"| Intervals
    JIT -.->|"loads on call"| Stitch

    style Gateway fill:#1a1a2e,color:#fff
    style Router fill:#00ADD8,color:#fff
    style JIT fill:#00ADD8,color:#fff
    style Firewall fill:#00ADD8,color:#fff
    style Pool fill:#00ADD8,color:#fff
    style Federation fill:#00ADD8,color:#fff

---

v0.7.0: What's New

| Feature | Description | |:--------|:------------| | 🔐 OAuth2 Authentication | Built-in support for HTTP MCP servers with OAuth2 | | 🔄 Streamable HTTP | Full Streamable HTTP transport implementation for MCP | | 👥 Hierarchical Namespaces | Multi-team organization with namespace assignment | | ⚡ Connection Pooling | HTTP clients with connection reuse and rate limiting | | 🧠 Lazy Schema Loading | Schemas load only when tools are actually called | | 🔧 Session Re-initialization | Fast session recovery without full restart | | 💰 Cost Attribution | Per-server cost tracking and reporting |

---

Join the Community

GitHub • Documentation • Issues

---

License

MIT © Marco Mornati

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: mmornati
• Source: mmornati/leanproxy-mcp
• License: MIT

Install and usage instructions live in the source repository linked above.