Motivation: My cofounder and I were tired of switching between pgAdmin-style GUIs and the psql CLI. Some tasks are faster in a UI, others in a terminal. So we built a TUI inspired by k9s that tries to cover both.

Features: * Visual WHERE condition builder (column / operator / value via dropdowns) * Natural language → SQL via AI (explicit consent required); supports OpenAI, Anthropic, Ollama (more planned) * Schema-aware autocomplete for tables and columns * Grid-based table browser with keyboard navigation * Built-in MCP server (we use it with Claude Code locally and want to add it to the official plugin list) * Supports all core WhoDB databases: Postgres, MySQL, MariaDB, SQLite, MongoDB, Redis, ClickHouse, Elasticsearch

Performance: * Startup: <300ms * Query execution: ~50–200ms overhead vs raw SQL (mostly rendering) * Binary size: ~50MB (OS/arch dependent)

Tradeoffs: 1. Optimized for interactive exploration, not bulk operations. Pagination exists but very large tables can be painful. 2. TUI complexity: Bubble Tea is powerful but harder to debug than a simple CLI. Errors are shown with copy-pasteable GitHub issue text. 3. AI requires setup: Ollama locally or env vars (WHODB_OPENAI_API_KEY, WHODB_ANTHROPIC_API_KEY). 4. Keyboard shortcuts may not feel intuitive to everyone—feedback welcome.

Known limitations: 1. Syntax highlighting is basic and sometimes wrong. 2. Large result sets can feel sluggish when scrolling, even with pagination. 3. Connections are file-based (~/.whodb-cli/config.yaml). Passwords/secrets are stored in the system keyring.

This tool was built primarily for ourselves as software developers. It’s not meant for heavy analytics or DBA-style power tooling, but we plan to keep improving it and make it more enterprise-ready.

CLI README with usage examples: [https://github.com/clidey/whodb/blob/main/cli/README.md](https://github.com/clidey/whodb/blob/main/cli/README.md)

Open questions: 1. Is the MCP server actually useful to others? 2. Is a simple yes/no AI consent sufficient, or is there a better/compliant approach? 3. Is tab-based view switching intuitive, or would vim-style :commands be better? 4. How should this fit into existing workflows instead of feeling “extra”?

Try it: npm i -g @clidey/whodb-cli whodb-cli

Supports macOS, Windows, and Linux (arm64/x64, plus armv7 without AI). We’ll add Homebrew and go install support. Binaries are also attached to releases: https://github.com/clidey/whodb/releases/tag/0.89.0