# Terragon Labs - Cloud-Based AI Coding Agents ## Overview Terragon Labs is a cloud-based platform that enables developers to run multiple AI coding agents in parallel to handle software development tasks asynchronously. Built as an orchestration layer for AI assistants like [[Claude Code]], OpenAI Codex, and Amp, it lets you delegate coding work that executes in isolated cloud sandboxes - accessible from your browser, mobile phone, or terminal. The platform automates everything from code generation and testing to pull request creation, allowing developers to work on 3-10 concurrent tasks simultaneously depending on their subscription tier. ## Key Features - **Parallel AI Agent Execution**: Run 3-10 concurrent coding tasks simultaneously without managing multiple worktrees - **Multiple AI Agent Support**: Compatible with Claude Code (including Sonnet 4.5), OpenAI Codex, and Amp - **Cloud Sandboxes**: Isolated development environments with 2 CPU cores and 4GB RAM (larger sizes coming to Pro plan) - **Deep GitHub Integration**: @mention bot in issues/PRs for context-aware task creation, automatic PR generation with co-author attribution - **Terry CLI**: Command-line interface (`@terragon-labs/cli`) for task management and pulling cloud-developed code locally - **Mobile-First Design**: Full task management capabilities from mobile devices during commutes or on-the-go - **Slack Integration**: Trigger tasks and receive notifications directly through Slack - **MCP Server Support**: Bridge between [[Claude Code]] and Terragon for natural language task management - **Automations**: Schedule recurring tasks or create event-triggered workflows for maintenance work - **Environment Customization**: Custom setup scripts (`terragon-setup.sh`), environment variables, and system prompts - **Built-in GitHub CLI**: `GH_TOKEN` automatically configured in sandboxes for repo context access - **Slash Commands**: `/clear` and `/compact` for agent control during task execution - **CI Integration**: Quick actions to fix failing checks and mark PRs ready for review ## Use Cases - **Parallel Development**: Work on multiple features, bug fixes, or refactorings simultaneously without complex git worktree management - **Mobile Development Workflows**: Start coding tasks from your phone during commutes, review PRs when you get home - **Automated Bug Fixes**: Delegate bug resolution to agents via GitHub issue @mentions with full context - **Code Review Assistance**: Ask agents to implement reviewer suggestions directly through PR comment @mentions - **Routine Maintenance**: Automate dependency updates, documentation generation, changelog writing, and other repetitive tasks - **Rapid Prototyping**: Spin up isolated environments to test new approaches without polluting your local development setup - **CI/CD Automation**: Auto-fix failing tests, update configurations, refactor code based on linter warnings - **Distributed Team Collaboration**: Enable asynchronous work delegation across time zones for remote teams - **Learning & Experimentation**: Safe sandboxes for trying new libraries or patterns without local setup overhead ## Platform Support - **Web Application**: Primary interface at terragonlabs.com with full task management capabilities - **Mobile**: Mobile-friendly web interface optimized for iOS and Android browsers - **Desktop**: Compatible with macOS, Windows, and Linux through web browser access - **CLI**: Terry CLI available via npm (`npm install -g @terragon-labs/cli`) for macOS, Linux, and WSL - **Integration**: MCP server integration for [[Claude Code]] users wanting natural language task management ## Pricing - **Free Trial**: 14-day trial with no credit card required to test all Core features - **Core Plan**: $25/month - 2 CPU, 4GB RAM sandboxes - 3 concurrent tasks - 5 automations - GitHub integration with @mention bot - Slack integration - Bring your own Claude Max or ChatGPT subscription - **Pro Plan**: $50/month - 2 CPU, 4GB RAM sandboxes (larger sizes coming soon) - 10 concurrent tasks - Unlimited automations - GitHub and Slack integrations - Bring your own AI subscriptions - REST API access (coming soon) - **Enterprise**: Custom pricing with contact sales - Bring Your Own Sandboxes - Team Automations and Team Environments - Custom integrations and advanced security ## Integrations ### Current Integrations - **GitHub**: Native integration with @mention bot, automatic PR creation, branch management, issue/PR context awareness, GitHub CLI access, co-author attribution - **Slack**: Task triggering and notification delivery through Slack workspace integration - **Claude Code**: MCP server integration supporting Sonnet 4.5, works with user's Claude Max subscription - **OpenAI Codex**: Integration supporting user's ChatGPT subscription and API keys - **Amp**: AI agent configuration via API key in Settings ### Coming Soon - **Linear**: Project management integration (mentioned in UI mockups) - **Jira**: Issue tracking integration (mentioned in UI mockups) - **REST API**: Custom integration capabilities for Pro plan users ## Installation ### Web Setup 1. Sign up at [terragonlabs.com](https://www.terragonlabs.com) 2. Connect your GitHub account and grant repository access permissions 3. Configure your AI agent (Claude Code, OpenAI Codex, or Amp) with API keys 4. Create your first task by selecting a repository and describing the work 5. Review the auto-generated pull request and merge when ready ### CLI Installation ```bash # Install Terry CLI globally npm install -g @terragon-labs/cli # Pull task changes to local repository terry pull ``` ### Requirements - GitHub account (required for repository access) - Claude Max or ChatGPT subscription (bring your own API keys) - Node.js and npm (for CLI installation) - Modern web browser for web interface access ### Optional Configuration - Create `terragon-setup.sh` in repository root for custom environment setup - Configure environment variables for project-specific needs - Set up custom system prompts for coding standards in Settings - Enable faster package managers (pnpm, uv) in setup scripts for quicker initialization ## Technical Details ### AI Models Supported - **Claude Code** (Anthropic) - including latest Sonnet 4.5 model - **OpenAI Codex** - GPT-4 based coding assistant - **Amp** - Alternative AI coding agent ### Sandbox Specifications - **Current**: 2 CPU cores, 4GB RAM per isolated environment - **Upcoming**: Larger sandboxes for Pro plan subscribers - **Isolation**: Full environment isolation per task to prevent conflicts - **Capabilities**: Install dependencies, run tests/builds/lint, fetch web resources, GitHub CLI access ### API Access - **Status**: Coming soon to Pro plan - **Type**: REST API for programmatic task management - **Use Cases**: Custom integrations, workflow automation, external tool connections ### CLI Tool - **Name**: Terry CLI - **Package**: `@terragon-labs/cli` - **Installation**: `npm install -g @terragon-labs/cli` - **Platforms**: macOS, Linux, WSL (Windows Subsystem for Linux) - **Primary Command**: `terry pull` - Pull task changes to local repository for testing ### Security Features - GitHub App with granular repository access permissions - Isolated sandboxes prevent cross-task contamination - GH_TOKEN customizable with specific permissions if needed - Bring-your-own API keys for AI models (no shared credentials) - Enterprise plan supports bring-your-own-sandbox for additional control - Co-author attribution on all PRs for transparency ## How Terragon Labs Differs from Claude Code For developers using [[Claude Code]] via API or Claude MAX subscription, understanding what Terragon actually adds is crucial. **Important**: claude.ai/code (Claude's web interface) already provides background task processing, parallel execution, automated PR creation, and mobile browser access. ### What claude.ai/code Already Does **Claude Code Web Interface** (claude.ai/code) includes: - ✅ **Background task processing** - Tasks run in cloud sandboxes, you can close browser - ✅ **Parallel execution** - Run multiple tasks simultaneously in isolated VMs - ✅ **Automated Git workflow** - Auto-creates branches, commits, and PRs via GitHub integration - ✅ **Mobile browser access** - Full functionality from mobile browsers - ✅ **GitHub App integration** - Connect repos, manage permissions, review PRs - ✅ **Cross-device continuity** - Start on desktop, continue on mobile, or vice versa ### What Terragon Actually Adds Given the overlap, Terragon's **unique value propositions** are: **1. GitHub @mention Bot** - Comment `@terragon-labs fix the failing tests` in GitHub issues/PR comments - Automatically creates task with full issue/PR context - claude.ai/code requires manually visiting website and creating task **2. Multi-Agent Support** - Run OpenAI Codex or Amp agents instead of Claude - Compare different AI approaches on same task - claude.ai/code only supports Claude models **3. Scheduled Automations** - Configure recurring tasks (e.g., "update dependencies every Monday") - Event-triggered workflows - claude.ai/code has no scheduling capabilities **4. Terry CLI Integration** - `terry pull` command for specific Terragon workflow - CLI-based task management for Terragon platform - Different from Claude's official CLI **5. Alternative UI/Dashboard** - Different task management interface than claude.ai/code - Centralized view across different AI agents - Cost tracking and analytics dashboard ### Honest Comparison | Feature | claude.ai/code | Terragon Labs | |---------|---------------|---------------| | **Background Processing** | ✅ Yes | ✅ Yes | | **Parallel Tasks** | ✅ Yes | ✅ Yes (3-10 depending on plan) | | **Auto PR Creation** | ✅ Yes | ✅ Yes | | **Mobile Access** | ✅ Yes (mobile browser) | ✅ Yes (mobile browser) | | **GitHub Comment Triggers** | ❌ No | ✅ Yes (@mentions) | | **Multi-Agent (Codex/Amp)** | ❌ No (Claude only) | ✅ Yes | | **Scheduled Tasks** | ❌ No | ✅ Yes | | **Cost** | Included in subscription | +$25-50/month extra | ### Cost Reality Check **Claude MAX Ultimate**: $200/month - Includes claude.ai/code web interface with all features - Background processing, parallel tasks, auto PRs already included - 20x extended usage limits **Claude MAX Ultimate + Terragon**: $225-250/month (+$25-50) - Same Claude features PLUS: - GitHub @mention bot integration - Multi-agent support (Codex/Amp comparison) - Scheduled/recurring automations - Alternative UI/dashboard **Is the extra $25-50/month worth it?** - **Probably not** if you're satisfied with claude.ai/code's native capabilities - **Maybe** if GitHub @mentions would save significant workflow friction - **Yes** if you need scheduled automations or want to compare Claude vs Codex approaches - **Yes** if managing multiple AI agents from one dashboard is valuable ### When to Use claude.ai/code (No Terragon) ✅ **claude.ai/code is sufficient** - Already has parallel tasks, background processing, auto PRs ✅ **Manual task creation is fine** - Visiting claude.ai/code to start tasks works for your workflow ✅ **Claude-only is enough** - Don't need to compare with Codex or other agents ✅ **No scheduled work** - Don't need recurring/automated task triggers ✅ **Budget-conscious** - Avoid extra $25-50/month on top of Claude subscription ### When Terragon Adds Real Value ✅ **GitHub comment workflows** - @mention bot to create tasks directly from issue/PR discussions saves time ✅ **Multi-agent comparison** - Want to test Claude vs OpenAI Codex vs Amp on same tasks ✅ **Scheduled automations** - Need recurring tasks like "update dependencies every Monday" ✅ **Unified dashboard** - Managing multiple AI agents across different providers ✅ **Team workflows** - Centralized task tracking for team using different AI models ### Practical Decision Framework **Use claude.ai/code (skip Terragon) if**: - You're happy with Claude's native web interface - Manually creating tasks doesn't bother you - You only use Claude (no need for Codex/Amp) - Cost-sensitive about the extra $25-50/month **Add Terragon only if**: - GitHub @mentions would genuinely improve your workflow - You want to compare multiple AI agents (Claude vs Codex) - Scheduled/recurring automations are critical - You value the specific Terragon UI/dashboard over claude.ai/code **Honest assessment**: For most Claude MAX users, **claude.ai/code alone is the better choice**. Terragon's $25-50/month premium is hard to justify unless GitHub @mentions or scheduled automations are genuinely valuable to your specific workflow. ## Best Practices ✅ **Break complex tasks into smaller, focused subtasks** for better agent performance and faster iteration ✅ **Use environment setup scripts** (`terragon-setup.sh`) to optimize dependency installation and reduce sandbox startup times ✅ **Enable faster package managers** (pnpm, uv) in setup scripts for quicker initialization ✅ **Leverage @terragon-labs mentions** in GitHub issues and PRs for context-aware task creation with full conversation history ✅ **Pull tasks locally** (`terry pull`) for final testing before merging to catch environment-specific issues ✅ **Configure custom system prompts** for project-specific coding standards and conventions ✅ **Review agent work before merging** - treat PRs as collaborative code review, not blind acceptance ✅ **Use automations** for recurring tasks like weekly dependency updates and changelog generation ✅ **Provide explicit details** in task prompts - agents work best with specific, well-defined instructions ## Limitations ⚠️ **Tasks too broad or vague** - Agents perform better with specific, focused work rather than open-ended exploration ⚠️ **Repository access required first** - Must grant GitHub repository access before @mentions work ⚠️ **Concurrent task limits** - 3 tasks on Core, 10 on Pro - additional tasks queue until slots open ⚠️ **No implicit context understanding** - Must provide explicit details in prompts; agents don't infer unstated requirements ⚠️ **Sandbox resource constraints** - Current 2 CPU/4GB RAM may be limiting for very large codebases (larger sizes coming) ⚠️ **Always verify locally** - Pull and test builds/tests locally before merging agent-generated PRs ⚠️ **Beta platform** - Currently in production beta with active development and feature additions ## Related Tools - [[Claude Code]] - AI coding assistant that integrates with Terragon via MCP server - Conductor.build - Alternative focusing on local git worktree management - Cursor - IDE-integrated AI coding assistant for local development - GitHub Copilot - IDE code completion and suggestion tool ## Resources - **Official Website**: [https://www.terragonlabs.com](https://www.terragonlabs.com) - **Documentation**: [https://docs.terragonlabs.com/docs](https://docs.terragonlabs.com/docs) - **GitHub Organization**: [https://github.com/terragon-labs](https://github.com/terragon-labs) - **Support**: [email protected] - **Terry CLI Package**: [@terragon-labs/cli on npm](https://www.npmjs.com/package/@terragon-labs/cli) ## Remarks Terragon Labs is a cloud orchestration platform for AI coding agents that adds GitHub @mention integration, multi-agent support (Claude/Codex/Amp), and scheduled automations on top of existing AI coding tools. For Claude MAX users specifically, the value proposition is narrow: claude.ai/code already provides background processing, parallel tasks, automated PR creation, and mobile access - all features Terragon initially marketed as differentiators. Terragon's actual unique value lies in workflow conveniences (GitHub comment-triggered tasks), multi-agent experimentation (comparing Claude vs Codex approaches), and scheduled recurring work (automated dependency updates). For most developers using Claude exclusively, the $25-50/month premium over claude.ai/code's included functionality is difficult to justify unless GitHub @mention workflows or scheduled automations are critical to their specific development process.