This topic surveys open-source scanners and developer tooling focused on AI safety, automated testing, and code-review workflows as they stand in mid‑2026. Demand for reproducible model testing, supply‑chain transparency, and privacy‑preserving code assistance has pushed teams toward self‑hosted scanners and developer SDKs that integrate model evaluation into CI/CD and governance pipelines. Key capabilities include static and dynamic scanning of prompts and model outputs, automated adversarial testing, observability for agent behavior, and local-first code completion for sensitive repositories. Representative projects reflect these priorities: LangChain provides a developer‑first framework and SDKs to build, observe, and deploy reliable LLM agents and embed testing and runtime guards; Tabby is an open‑source, self‑hosted coding assistant with IDE extensions and model serving for private, low‑latency completions; CodeGeeX and Salesforce CodeT5 (CodeT5+) are open code models for generation and understanding used in automated code review and completion; WizardLM derivatives offer instruction‑tuned models adapted for developer workflows; Stable Code delivers edge‑ready code models optimized for fast, private completion. Perplexity’s Bumblebee (and like‑minded scanners) exemplifies the growing class of dedicated safety scanners that automatically detect risky outputs, prompt injection, and policy violations as part of pre‑merge and runtime checks. For teams building governance and test automation, the practical choice today is combining model‑aware scanning (Bumblebee‑style) with developer platforms (LangChain, Tabby) and open code models to retain control over data, reproducibility, and audit trails while scaling code review and safety testing.