I built CALEA because I was frustrated with the constant context-switching required to test modern applications: writing Pytest tests, configuring Selenium, running OWASP ZAP via Docker, checking Lighthouse, and more.
CALEA is a desktop application (packaged with Nuitka) that acts as an autonomous agent. You give it a high-level goal — for example, “Audit the security of this API” — and it orchestrates the underlying tools to generate reports.
Under the hood, the stack is based on Python 3.11, with FastAPI handling internal asynchronous logic and PyWebview used for the UI. The architecture relies on the Model Context Protocol (MCP) to expose local tools such as the file system, terminal, and browser to the LLM.
On the AI side, CALEA supports cloud models like GPT-4 and Claude, but it is optimized for local execution via Ollama for privacy reasons — Qwen 2.5 being my personal favorite.
It’s a paid product, built as a bootstrapped solo project, but there’s a 7-day free trial (no card required) for anyone who wants to try it out. I’d love to hear your feedback, whether on the architecture or the tool itself.
Hi HN, I’m the author.
I built CALEA because I was frustrated with the constant context-switching required to test modern applications: writing Pytest tests, configuring Selenium, running OWASP ZAP via Docker, checking Lighthouse, and more.
CALEA is a desktop application (packaged with Nuitka) that acts as an autonomous agent. You give it a high-level goal — for example, “Audit the security of this API” — and it orchestrates the underlying tools to generate reports.
Under the hood, the stack is based on Python 3.11, with FastAPI handling internal asynchronous logic and PyWebview used for the UI. The architecture relies on the Model Context Protocol (MCP) to expose local tools such as the file system, terminal, and browser to the LLM.
On the AI side, CALEA supports cloud models like GPT-4 and Claude, but it is optimized for local execution via Ollama for privacy reasons — Qwen 2.5 being my personal favorite.
It’s a paid product, built as a bootstrapped solo project, but there’s a 7-day free trial (no card required) for anyone who wants to try it out. I’d love to hear your feedback, whether on the architecture or the tool itself.