HKhkdeman·Jan 20, 2026Submission
Show HN: WhoDB CLI – Terminal database client (Golang) with local AI support
We built a terminal database client that lets you connect to multiple databases and interact with them through a TUI. It’s open source (Apache 2.0).
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, Red
Read the rest
is, 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