Bloop started as a desktop AI-powered code search engine that let developers query their codebases in natural language. The tool, written in Rust, offered semantic search using on-device MiniLM embeddings, regex search, go-to-reference/definition navigation for 10+ languages via Tree-sitter, and a GPT-4-backed conversational interface. It was available free for Mac, Windows, and Linux and kept all indexing local for privacy. The original bloop code search repository was archived by the BloopAI organization on January 2, 2025, marking the end of active development on that product. The company, founded in 2021 by Louis Knight-Webb and Gabriel Gordon-Hall and backed by Y Combinator with $7.43M raised, subsequently pivoted its focus toward enterprise legacy code modernization. Bloop's current primary offering is ModerniseAI, which uses AI to convert legacy COBOL codebases into readable, maintainable Java. The process starts with a system survey that risk-scores each component and its dependencies, proceeds through a proof-of-concept sprint, and then performs full system modernization. Built-in correctness checks are used to validate functional equivalence between the original COBOL and the generated Java. The company targets large enterprises with significant mainframe and COBOL estates that face pressure to modernize. The approach positions itself against manual rewriting projects by using AI to reduce time, cost, and human error. Bloop maintains a dedicated research lab focused on AI and programming languages to underpin the translation quality. As of mid-2026, the BloopAI GitHub organization remains active with several projects including a vibe-kanban tool and dev-manager-mcp, though the original code search product is archived and no longer maintained. Key features: - COBOL-to-Java automated translation with functional equivalence guarantees - Risk-scored system survey to prioritize modernization of legacy components - Proof-of-concept sprint before full system modernization commitment - Built-in correctness checks to validate translated code against original behavior - Original code search product offered on-device MiniLM semantic embeddings (now archived) - Original product supported GitHub, GitLab, Bitbucket, and local folders - Tree-sitter-based go-to-reference and go-to-definition for 10+ languages (legacy product)
Not public (ModerniseAI enterprise pricing not disclosed; original desktop app was free)