Developer Martin Schenk launches privacy-first terminal that decides where your data goes before making any call to ChatGPT
Published: November 2025
The Privacy Paradox
In an era where developers increasingly rely on AI assistants like ChatGPT and Claude for coding help, a critical question has emerged: how much of our sensitive data are we inadvertently exposing to cloud services?
Martin Schenk, a developer based in Spain, faced this concern firsthand after accidentally pasting API credentials into ChatGPT during a debugging session. The incident sparked the development of AI Chat Terminal, an open-source tool that addresses a problem many developers face but few openly discuss.
«Every time I use ChatGPT for coding help, I’m sending my entire conversation history to their servers,» Schenk explains. «Passwords, API keys, database credentials, client information—it’s all logged. The convenience is undeniable, but the privacy cost is high.»
The tool Schenk built takes a different approach: instead of routing all queries to cloud-based AI services, AI Chat Terminal analyzes user input for sensitive keywords before any network call is made. If sensitive data is detected, the query is automatically routed to a local AI model running on the user’s Mac. Otherwise, it leverages the full power of cloud-based models like GPT-4o.
How It Works
The system’s architecture is surprisingly straightforward. When a user types a command, the terminal performs a sub-millisecond keyword scan. Commands containing words like «save,» «password,» «API key,» or similar sensitive terms trigger local processing.
For example:
- «save my database password xyz123» → Processed locally by Qwen 2.5 Coder AI, encrypted with AES-256, stored in SQLite. Never touches the cloud.
- «what’s the best way to parse JSON in Python?» → Sent to OpenAI’s GPT-4o for a fast, comprehensive answer.
The keyword detection system supports 3 languages, including English, German, Spanish, making it accessible to developers worldwide.
Technical Stack
AI Chat Terminal is built on a combination of proven technologies:
- Local AI: Qwen 2.5 Coder (7B parameters) running via Ollama
- Encryption: SQLCipher with AES-256 encryption for sensitive data storage
- Cloud AI: OpenAI GPT-4o for general queries
- Architecture: Python + Zsh with a daemon-based design for fast response times
- Security: macOS Keychain integration for API key management
The tool is currently macOS-exclusive, though Schenk is actively seeking contributors to port it to Linux and Windows.
Privacy-First Design Philosophy
What sets AI Chat Terminal apart from other AI wrapper tools is its privacy-first design decisions:
- No persistent conversation logs: Chat history is automatically deleted on exit
- Local-first processing: Sensitive data never leaves the user’s machine
- Transparent routing: Users control which data goes where through simple keywords
- Open source: The entire codebase is auditable (MIT License)
- Zero telemetry: No usage tracking or analytics
«I wanted complete transparency,» Schenk notes. «You should know exactly where your data is going at all times. There are no hidden cloud syncs, no telemetry, no tracking.»
v11.6.2: Intelligent Keyword Matching
The latest release (v11.6.2) introduces intelligent multi-keyword matching that understands natural language context. The system can now detect compound sensitive patterns like:
- «Please save this API key for production»
- «Store my client’s database credentials»
- «Remember this OAuth token»
The improved matching algorithm reduces false positives while maintaining security, ensuring that genuinely sensitive data is consistently routed to local processing.
Installation: 30 Seconds
True to the developer ethos of simplicity, installation is a one-line command:
curl -fsSL https://raw.githubusercontent.com/martinschenk/ai-chat-terminal/main/install.sh | zshThe installer automatically sets up Ollama, Qwen 2.5 Coder model, SQLCipher encrypted database, background daemon, and terminal integration.
Requirements are minimal: macOS 12.0+, Python 3.9+, and an OpenAI API key.
Use Cases
AI Chat Terminal is designed for:
- Developers who chat with AI daily but handle sensitive code
- DevOps engineers managing credentials and infrastructure
- Security-conscious professionals who can’t risk cloud exposure
- Anyone who has accidentally pasted something sensitive into ChatGPT
Industry Context
AI Chat Terminal emerges at a critical moment in developer tooling. As AI coding assistants become ubiquitous—with GitHub Copilot, Cursor, and Claude integrations standard in modern development workflows—the question of data privacy grows more urgent.
Recent surveys show that over 70% of developers now use AI assistants regularly, yet many companies prohibit their use due to data leakage concerns. Tools like AI Chat Terminal offer a potential middle ground: the productivity benefits of AI without the privacy trade-offs.
Open Source and Community
The project is fully open source under the MIT License, hosted on GitHub at github.com/martinschenk/ai-chat-terminal. Schenk actively welcomes contributions, particularly:
- Linux and Windows ports
- Additional AI backend integrations (Anthropic Claude, local LLMs)
- Security audits
- Documentation improvements
The project’s landing page at ai-chat-terminal.martin-schenk.es provides comprehensive documentation, architecture details, and usage examples.
The Road Ahead
Schenk’s vision for AI Chat Terminal extends beyond macOS. «The core problem—balancing AI convenience with data privacy—is universal,» he explains. «I’d love to see this work on Linux, Windows, and eventually integrated into other tools.»
Future development priorities include cross-platform support, additional AI backends, IDE integrations (VS Code, JetBrains), team/enterprise features, and mobile support.
Conclusion
AI Chat Terminal represents a pragmatic approach to a growing problem in software development. Rather than choosing between AI productivity and data privacy, developers can now have both—with a simple keyword determining the routing.
For developers concerned about inadvertently exposing sensitive data while still wanting AI assistance, Schenk’s tool offers a compelling solution. And with its open-source nature, the community can verify, audit, and improve the privacy guarantees over time.
As AI becomes increasingly integrated into developer workflows, tools like AI Chat Terminal may represent the future: intelligent systems that respect privacy by design, not as an afterthought.
About the Creator
Martin Schenk is a developer and entrepreneur based in Spain. His main project is Mindbeamer, an AI-powered bookmark manager that helps users organize and rediscover their saved content. He also maintains Startups Españolas, a directory of Spanish startups. AI Chat Terminal is his latest contribution to the developer tools ecosystem.
Learn more at martin-schenk.es
Project Links
- Website: ai-chat-terminal.martin-schenk.es
- GitHub: github.com/martinschenk/ai-chat-terminal
- License: MIT (Open Source)
Editor’s Note: This article was written for Startups Españolas as part of our coverage of innovative developer tools emerging from the Spanish tech ecosystem.