Nativelink MCP Server

MCP (Model Context Protocol) server for Nativelink Cloud - High-performance cloud build cache and remote execution system that accelerates Bazel builds. Sign up at app.nativelink.com to get your personalized configuration.

Features

• 🚀 Generate optimized Bazel configurations using Nativelink Cloud endpoints
• 📚 Access Nativelink Cloud setup and migration guides directly in your AI assistant
• 📊 Analyze build performance with AI-powered recommendations
• 🔄 Set up automatic builds and tests with file watchers and Nativelink caching
• 🏗️ Generate deployment configurations for self-hosted Nativelink instances
• 🔐 Integration with app.nativelink.com for personalized configurations
• 💡 Smart recommendations based on your project's specific needs

Installation

Quick Install (Cursor)

Or manually add to your Cursor configuration (~/.cursor/mcp.json):

{
  "mcpServers": {
    "nativelink": {
      "command": "npx",
      "args": ["-y", "@nativelink/mcp-server", "--api-key", "YOUR_API_KEY"]
    }
  }
}

Claude Code

claude mcp add nativelink -- npx -y @nativelink/mcp-server --api-key YOUR_API_KEY

VS Code

Add to your VS Code MCP settings:

{
  "servers": {
    "nativelink": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@nativelink/mcp-server", "--api-key", "YOUR_API_KEY"]
    }
  }
}

Getting Started with Nativelink Cloud

1. Sign up at app.nativelink.com

• Create your free account
• Get instant access to cloud build cache and remote execution

2. Get Your Personalized Configuration

After signing up, you'll receive a personalized .bazelrc configuration:

build --remote_cache=grpcs://cas-tracemachina-shared.build-faster.nativelink.net
build --remote_header=x-nativelink-api-key=YOUR_API_KEY
build --bes_backend=grpcs://bes-tracemachina-shared.build-faster.nativelink.net
build --bes_header=x-nativelink-api-key=YOUR_BES_API_KEY
build --bes_results_url=https://app.nativelink.com/a/YOUR_BUILD_ID/build
build --remote_executor=grpcs://scheduler-tracemachina-shared.build-faster.nativelink.net:443

3. Add to Your Project

Append the configuration to your .bazelrc file and start building!

AI Provider Keys (Optional)

For enhanced AI-powered features:

• Anthropic: Get from console.anthropic.com
• Gemini: Get from makersuite.google.com

Usage

Simply add use nativelink to your prompts:

Set up Nativelink remote cache for my Rust project. use nativelink

Debug why my Bazel builds are slow. use nativelink

Set up automatic builds and tests after file changes. use nativelink

Available Tools

1. get-bazel-config

Generates optimized Bazel configuration for your project.

Parameters:

• projectType (required): rust, cpp, java, python, go, or mixed
• nativelinkUrl (optional): Custom Nativelink server URL
• features (optional): Array of features to enable

Example:

Generate a Bazel config for my Rust project with remote cache and execution. use nativelink

2. get-nativelink-docs

Fetches relevant Nativelink documentation.

Parameters:

• topic (required): setup, migration, optimization, troubleshooting, or api
• context (optional): Additional context for the query
• maxTokens (optional): Maximum response length

Example:

How do I migrate from local builds to Nativelink cloud? use nativelink

3. analyze-build-performance

Analyzes build metrics and provides optimization recommendations.

Parameters:

• profileData (optional): Bazel profile JSON
• metrics (optional): Build performance metrics
• targetOptimization (optional): speed, cost, or balanced

Example:

My cache hit rate is 40%. How can I improve it? use nativelink

4. generate-deployment-config

Generates deployment configurations for various platforms.

Parameters:

• platform (required): kubernetes, docker, aws, gcp, or azure
• scale (required): small, medium, large, or enterprise
• features (optional): monitoring, autoscaling, high_availability

Example:

Generate a Kubernetes deployment for Nativelink with autoscaling. use nativelink

5. setup-watch-and-build

Sets up automatic Bazel builds and tests on file changes with Nativelink Cloud caching.

Parameters:

• command (optional): build, test, or both (default: both)
• targets (optional): Bazel targets to build/test (default: //...)
• useIbazel (optional): Use iBazel for intelligent watching (recommended)
• watchPaths (optional): Paths/patterns to watch
• excludePaths (optional): Paths/patterns to exclude
• debounceMs (optional): Delay before rebuilding (default: 1000ms)

Example:

Set up automatic builds and tests when files change. use nativelink

What it provides:

• Configuration for iBazel (intelligent Bazel watcher)
• Setup instructions for watchexec, entr, nodemon
• Custom watch scripts
• VS Code task configuration
• Integration with Nativelink Cloud for instant cache hits

Configuration

Environment Variables

export NATIVELINK_API_KEY=your_api_key
export ANTHROPIC_API_KEY=your_anthropic_key  # Optional
export GEMINI_API_KEY=your_gemini_key        # Optional
export NATIVELINK_URL=https://custom.url     # Optional

Command Line Arguments

npx @nativelink/mcp-server \
  --api-key YOUR_API_KEY \
  --anthropic-key YOUR_ANTHROPIC_KEY \
  --transport http \
  --port 8080

Pro Tips

Add a Rule

To avoid typing use nativelink every time, add this rule to your AI assistant:

Cursor: Add to Settings > Rules Claude Code: Add to CLAUDE.md

Always use nativelink when I ask about:
- Bazel configuration or optimization
- Build cache setup
- Remote execution
- Build performance issues

Project-Specific Configuration

Create .bazelrc.nativelink in your project:

# Generated by Nativelink MCP
build --remote_cache=grpc://cache.nativelink.com:443
build --remote_cache_header=x-nativelink-api-key=YOUR_KEY
build --remote_download_minimal
build --experimental_remote_cache_compression

Then import in your main .bazelrc:

try-import .bazelrc.nativelink

Development

Building from Source

git clone https://github.com/nativelink/mcp-server-nativelink.git
cd mcp-server-nativelink
npm install
npm run build

Running Locally

# Stdio transport (for testing)
npm run dev

# HTTP transport
npm run dev -- --transport http --port 3000

Testing with MCP Inspector

npx @modelcontextprotocol/inspector npx @nativelink/mcp-server

Troubleshooting

Authentication Errors

• Verify API key is set correctly
• Check network connectivity
• Ensure firewall allows outbound HTTPS

Tool Not Working

• Update to latest version: npm update @nativelink/mcp-server
• Check logs with DEBUG=true environment variable
• Verify MCP client configuration

Performance Issues

• Use --remote_download_minimal for large artifacts
• Enable compression with --experimental_remote_cache_compression
• Check network latency to cache server

Support

• 📚 Nativelink Documentation
• 💬 Slack Community
• 🐛 GitHub Issues
• 📧 Enterprise Support: support@nativelink.com

License

FSL-1.1-Apache-2.0 - Functional Source License v1.1 with Apache 2.0 future license.

This software is licensed under the Functional Source License (FSL) v1.1 and will automatically convert to Apache 2.0 license two years after each release. See LICENSE file for full details.

Key Points:

• ✅ Free for internal use, education, and research
• ✅ Free for providing professional services to users of the software
• ⚠️ Cannot be used to create competing MCP servers or services
• 🔄 Converts to Apache 2.0 after 2 years

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Acknowledgments

Inspired by the excellent Context7 MCP Server architecture.