Kachilu Browser: A local browser automation CLI for AI agents

Kachilu Browser addresses a critical pain point in advanced AI agent development: reliable, private browser interaction. Standard automation often struggles with modern anti-bot measures, requiring cloud-based, and thus often invasive, relay services. Kachilu provides a robust, CLI-based solution that operates locally, making it ideal for developers prioritizing data sovereignty and stable, enterprise-grade automation. The technology's most valuable feature is its ability to detect and manage CAPTCHA-like challenges. When an agent encounters a reCAPTCHA v2/v3 or Cloudflare Turnstile, the browser doesn't fail; the challenge is detected, and completion is gracefully routed through the local CLI flow. This level of native, proactive challenge handling moves it far beyond simple web scraping tools and into the domain of genuine, resilient automation. From a developer workflow standpoint, Kachilu is highly engineered for integration with existing AI ecosystems. It is specifically designed to work with agent workflows (like the MCP prepare/exec cycle), ensuring that when context is compacted or sessions resume, the control plane, profile, and WSL2 Windows-browser target are preserved. The emphasis on WSL2 compatibility is particularly noteworthy, providing a seamless bridge for agents running in Linux containers to control the primary Windows user browser profile. While the tooling is advanced, the underlying philosophy—'Local by design. No hosted relay, no telemetry'—is a massive selling point for organizations handling sensitive data. For AI developers, this means building reliable, secure agents without compromising on privacy or incurring vendor lock-in associated with external, hosted services. The complexity of setup (dealing with WSL2 networking, path management, and specific environment variables) suggests a mature, albeit technically demanding, toolset.