Lobster

Meet Lobster, an AI assistant that can use your CLI and Editor.

Lobster is an advanced AI-powered development assistant designed to handle complex software tasks step-by-step. With tools that let it create & edit files, explore large projects, use the browser, and execute terminal commands (with your approval), Lobster goes far beyond simple code completion.

Unlike autonomous AI scripts that run in isolated sandboxes, Lobster provides a human-in-the-loop GUI, ensuring you approve every file change and terminal command for safety and transparency.

🚀 How Lobster Works

1. Enter your task and optionally add images (mockups, screenshots, diagrams).
2. Lobster analyzes your file structure, source code ASTs, runs regex searches, and reads relevant files.
3. Once ready, Lobster can:
   • Create and edit files while monitoring lint/compiler errors.
   • Execute terminal commands and react to output in real time.
   • Launch a browser for runtime debugging and automated testing.
4. When complete, Lobster presents final results and suggested commands.

🔌 Use Any API and Model

Lobster supports multiple API providers including:

• OpenAI
• Anthropic
• Google Gemini
• AWS Bedrock
• Azure
• GCP Vertex
• OpenRouter
• Cerebras
• Groq
• OpenAI-compatible APIs
• Local models via LM Studio / Ollama

Lobster tracks total tokens and estimated API costs per task, keeping usage transparent.

💻 Run Commands in Terminal

Lobster can execute commands directly in your terminal and monitor their output.

It can:

• Install dependencies
• Run build scripts
• Deploy applications
• Execute tests
• Monitor dev servers

Use "Proceed While Running" to let Lobster continue working while long-running processes execute in the background.

📝 Create and Edit Files

Lobster:

• Creates and edits files via diff view
• Lets you review, modify, or revert changes
• Monitors lint/compiler errors automatically
• Tracks all changes in VS Code Timeline

You stay in control at all times.

🌐 Use the Browser

Lobster can:

• Launch a browser
• Click, type, scroll
• Capture screenshots
• Monitor console logs

This enables automated UI testing, runtime debugging, and visual bug fixing.

Try:

“Test the app”

Lobster can start your dev server and validate functionality automatically.

🛠 Model Context Protocol (MCP)

Lobster supports the Model Context Protocol (MCP), allowing custom tools and servers.

You can:

• Install community MCP servers
• Create custom MCP tools
• Extend Lobster’s capabilities

Examples:

• “Add a tool that fetches Jira tickets”
• “Add a tool that manages AWS EC2 instances”
• “Add a tool that pulls PagerDuty incidents”

📎 Add Context with Mentions

Use special mentions to improve results:

• @url – Fetch and include webpage content
• @problems – Add workspace errors
• @file – Include file contents
• @folder – Include entire folders

🔄 Checkpoints: Compare & Restore

Lobster automatically snapshots your workspace during tasks.

You can:

• Compare changes
• Restore workspace only
• Restore full task + workspace

This enables safe experimentation without losing progress.

🤝 Contributing

We welcome contributions!

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

Repository:
https://github.com/NIHAR-SARKAR/Lobster

🏢 Enterprise (Future Roadmap)

Planned enterprise capabilities include:

• SSO (SAML/OIDC)
• Global policy controls
• Audit logging
• Private networking
• On-prem deployment options

👨‍💻 Developer

NIHAR-SARKAR
GitHub: https://github.com/NIHAR-SARKAR

📜 License

Apache 2.0 © 2026 NIHAR-SARKAR