Claude Code Restored Runtime

claude-code-restored-runtime

Skill by ara.so — Claude Code Skills collection.

What It Does

claude-code-rev is a restored and runnable version of Claude Code reconstructed from source maps. It provides:

• A local AI coding agent runtime that works with GitHub-backed inference providers
• Support for both GitHub Models (OpenAI-compatible) and GitHub Copilot endpoints
• Multi-provider switching with interactive CLI commands
• Chrome MCP and Computer Use MCP compatibility layers
• Restored CLI bootstrap with full command tree

The project is not the original upstream repository—some modules are shims or degraded implementations to enable runnable workflows.

Installation

Prerequisites

• Bun 1.3.5 or newer
• Node.js 24 or newer
• GitHub CLI (gh) for account-based authentication (optional but recommended)

Clone and Install

git clone https://github.com/oboard/claude-code-rev.git
cd claude-code-rev
bun install

Verify Installation

bun run version
bun run dev --help

Authentication Setup

Both GitHub providers use this lookup order:

1. Provider-specific environment variable
2. GH_TOKEN
3. GITHUB_TOKEN
4. gh auth token (GitHub CLI)

Option 1: GitHub CLI (Recommended)

gh auth login

This enables account-based authentication without manual token management.

Option 2: Environment Variable

export GITHUB_TOKEN="ghp_your_token_here"

Or add to .env:

GITHUB_TOKEN=ghp_your_token_here

Running the Runtime

Default Run

bun run dev

With GitHub Models Provider

bun run dev --settings '{"provider":"github-models"}'

Specify a model:

bun run dev --settings '{"provider":"github-models"}' --model "openai/gpt-4.1"

With GitHub Copilot Provider

bun run dev --settings '{"provider":"github-copilot"}'

Specify a Copilot-hosted Claude model:

bun run dev --settings '{"provider":"github-copilot"}' --model "claude-opus-4.6"

CLI Commands

Provider Management

Model Management

General Commands

bun run dev --help    # Show full command tree
bun run version       # Print restored version

Supported Models

GitHub Copilot – Validated Claude Models

These models work with the Claude Code runtime and tool loop:

• claude-sonnet-4.6
• claude-opus-4.6
• claude-haiku-4.5
• claude-sonnet-4.5
• claude-opus-4.5
• claude-sonnet-4

Note: Copilot GPT/Grok models are not fully wired yet—they require the /responses API path instead of the current chat/messages shim.

GitHub Models

Any OpenAI-compatible model available through GitHub Models, including:

• openai/gpt-4.1
• openai/gpt-4o
• meta-llama/Llama-3.3-70B-Instruct

Configuration Examples

Runtime Settings File

Settings are stored in user config and can be overridden via CLI:

# Inline settings override
bun run dev --settings '{"provider":"github-copilot","autoSave":true}'

Environment-Based Provider Override

# Force GitHub Models provider via env
export CLAUDE_CODE_PROVIDER="github-models"
bun run dev

Multi-Provider Workflow

# Start with GitHub Models
bun run dev --settings '{"provider":"github-models"}'

# Inside CLI, switch to Copilot
/provider github-copilot

# Pick a specific model
/model
# (select claude-opus-4.6 from picker)

Code Integration Examples

TypeScript: Provider Interface

import { GitHubModelsProvider } from './providers/github-models';

const provider = new GitHubModelsProvider({
  apiKey: process.env.GITHUB_TOKEN,
  model: 'openai/gpt-4.1',
});

const response = await provider.chat({
  messages: [
    { role: 'user', content: 'Explain this function' },
  ],
});

console.log(response.content);

TypeScript: Copilot Provider

import { GitHubCopilotProvider } from './providers/github-copilot';

const copilot = new GitHubCopilotProvider({
  token: process.env.GITHUB_TOKEN,
  model: 'claude-sonnet-4.6',
});

const result = await copilot.complete({
  prompt: 'Write a binary search in Rust',
  maxTokens: 500,
});

TypeScript: Runtime Bootstrap

import { bootstrap } from './cli/bootstrap';

async function main() {
  await bootstrap({
    provider: process.env.CLAUDE_CODE_PROVIDER || 'github-models',
    model: process.env.CLAUDE_CODE_MODEL,
    autoSave: true,
  });
}

main().catch(console.error);

Common Patterns

Pattern 1: Quick Provider Switch

# Start session
bun run dev

# Switch provider mid-session
/provider
# Select github-copilot from picker

# Verify active provider
/provider info

Pattern 2: Scripted Model Selection

#!/bin/bash
export GITHUB_TOKEN=$(gh auth token)
bun run dev \
  --settings '{"provider":"github-copilot"}' \
  --model "claude-opus-4.6"

Pattern 3: Fallback Chain

// Automatic fallback if primary provider fails
const providers = [
  new GitHubCopilotProvider({ token: process.env.GITHUB_TOKEN }),
  new GitHubModelsProvider({ apiKey: process.env.GH_TOKEN }),
];

for (const provider of providers) {
  try {
    return await provider.chat(messages);
  } catch (err) {
    console.warn(`Provider ${provider.name} failed:`, err);
  }
}
throw new Error('All providers failed');

Troubleshooting

Issue: bun install Fails

Cause: Bun version too old or missing native dependencies.

Fix:

bun --version  # Must be 1.3.5+
bun upgrade
bun install --force

Issue: Authentication Fails

Cause: No valid GitHub token found.

Fix:

# Verify token presence
gh auth status

# Re-authenticate
gh auth login

# Or manually set token
export GITHUB_TOKEN="ghp_..."

Issue: Model Not Available

Cause: Model name mismatch or provider doesn't support the model.

Fix:

# Use /model picker instead of manual entry
bun run dev --settings '{"provider":"github-copilot"}'
# Then run: /model

Issue: Copilot GPT Models Not Working

Cause: Copilot GPT/Grok models require /responses API (not yet fully wired).

Fix: Use Copilot-hosted Claude models instead:

bun run dev \
  --settings '{"provider":"github-copilot"}' \
  --model "claude-sonnet-4.6"

Issue: Source Map Restoration Artifacts

Cause: Some modules are shims or degraded implementations.

Symptoms: Missing features, unexpected behavior in non-core paths.

Fix: Check if the feature relies on a private/native integration. File an issue on the repo if critical functionality is broken.

Issue: Tool Loop Hangs

Cause: MCP compatibility layer returning malformed responses.

Fix:

# Enable debug logging
DEBUG=* bun run dev

# Check for MCP tool catalog errors
# Report specific tool failures to repo issues

Development Workflow

Run in Dev Mode

bun run dev

Inspect Restored CLI Structure

bun run dev --help
# Shows full command tree from restored bootstrap

Test Provider Switching

// In your test suite
import { switchProvider } from './cli/commands/provider';

await switchProvider('github-copilot');
const info = await getProviderInfo();
assert(info.active === 'github-copilot');

Key Files

• cli/bootstrap.ts – Restored CLI entry point
• providers/github-models.ts – GitHub Models provider
• providers/github-copilot.ts – GitHub Copilot provider
• mcp/chrome-mcp.ts – Chrome MCP compatibility layer
• mcp/computer-use-mcp.ts – Computer Use MCP shim

Limitations

• Not the original upstream codebase—some modules are restored from source maps
• Copilot GPT/Grok models not fully integrated (require /responses API)
• Private/native integrations replaced with shims
• Some dynamic imports and resource files may be incomplete

Resources

• Repository: https://github.com/oboard/claude-code-rev
• Provider Docs: Run /provider info in CLI for active provider details
• Model Catalog: Run /model to see available models for current provider