Skip to main content

Amp

Amp is a VSCode extension (available at https://ampcode.com) that provides AI-powered coding assistance with support for multiple language models.

Overview

Amp integrates directly into Visual Studio Code, providing coding assistance with a unique “Oracle” tool system that allows it to consult different language models for specialized tasks.

Multi-Model Support

Amp supports multiple language models, each with specialized tuning:

Claude Sonnet 4.x

Primary model with system prompt tuned specifically for Claude Sonnet 4.x

GPT-5

Available as alternative model with specialized tuning To use GPT-5, configure VSCode user settings:
{
    "amp.url": "https://ampcode.com/",
    "amp.gpt5": true
}

Other Models

Amp registers other LLMs as tools (“the Oracle”) that can be consulted for specific tasks.

Oracle Tool System

Unique feature where different language models are registered as tools:
  • Main model can consult other models for specialized tasks
  • Models act as “oracles” for specific types of queries
  • Allows leveraging different model strengths within single workflow

Accessing System Prompt

To obtain the system prompt for Amp:
  1. Login with Amp using VSCode
  2. Issue a short query into Amp
  3. Hold down Alt (Windows) or Option (macOS)
  4. Click on the workspace button
  5. Click “View Thread YAML”

Configuration

User Settings

Configure via VSCode user settings:
{
    "amp.url": "https://ampcode.com/",
    "amp.gpt5": true  // Enable GPT-5 mode
}

Model-Specific Tuning

Each supported model has:
  • Specialized system prompt tuning
  • Optimized for specific model capabilities
  • Different prompts for different models

Available Configurations

  • claude-4-sonnet.yaml: Tuned for Claude Sonnet 4.x
  • gpt-5.yaml: Tuned for GPT-5

Integration

VSCode Integration

  • Native VSCode extension
  • Workspace-aware
  • Thread-based conversations
  • YAML-formatted thread viewing

Thread Management

  • View conversation threads in YAML format
  • Track model interactions
  • Debug and understand AI decision-making

Key Features

  • Multi-model support with easy switching
  • Oracle tool system for consulting different models
  • Model-specific prompt tuning
  • Thread inspection capabilities
  • Workspace integration
  • Configurable via VSCode settings

Use Cases

  • Coding assistance with multiple model perspectives
  • Consulting different models for specialized tasks
  • Debugging with thread inspection
  • Model comparison within single workflow
  • Leveraging different model strengths

Build docs developers (and LLMs) love