Qazimut logoqazimut
Qazimut logo
Qazimut
v0.3.1 — now with C2PA stripping

Your AI dev co-pilot,in the terminal.

Compress images. Query databases. Learn faster. Generate commits. Explain code. Validate product ideas via Qyntel. All from one binary, one config, and pluggable AI providers — local-first for privacy.

$npx qazimut

macOS · Linux · Windows (WSL) · Docker

qazimut — learnzsh — 80×24
Live demo · qazimut learn "explain database indexes"
utility belt
Six tools you already reach for, in one binary.
Qazimut isn't another AI chat in the terminal. It's a coherent set of daily-driver tools — and the chat command ties them together via natural language.
qazimut chat
command
Talk to your tools in natural language.
qazimut chat "compress the images in ./public"
View docs
qazimut learn
command
Adaptive tutor with spaced-repetition review.
qazimut learn "explain database indexes"
View docs
qazimut img
command
Compress AI images, strip metadata (incl. C2PA).
qazimut img compress ./assets --preset web
View docs
qazimut db
command
Query databases in plain English. Read-only by default.
qazimut db query "top users last week"
View docs
qazimut commit
command
Conventional commit messages from your staged diff.
qazimut commit --type feat --scope auth
View docs
qazimut explain
command
Explain code, errors, and logs. Reads pipes & files.
cat error.log | qazimut explain --error
View docs
bring your own brain
Works with your stack.
Local-first by default. Bring your own key for the remote providers, or run Ollama for zero-cost, fully-offline inference.
  • OpenAIGPT-4, GPT-4o, o1
  • AnthropicClaude family
  • GoogleGemini family
  • OllamaLocal, default for privacylocal · default
  • LM StudioOpenAI-compatible local
  • llama.cppDirect llama.cpp
  • OpenRouterAggregator
Local-first. Bring your own key. Or use ours.
install
One binary. Five ways in.
Pick your poison. The npm global install is the happy path; the curl one-liner is the fastest; the rest are platform-native.
Install globally with npm.
npm i -g qazimut
$ after install, run qazimut setup to pick a provider (Ollama recommended), then try qazimut chat.
Currently showing: npm — Install globally with npm.
Qyntel
Validate your next product idea — from the terminal.
qazimut qyntel is a thin client that calls Qnaku's evidence-led validation engine — the same one behind the Qyntel web app. It runs the Problem → Workaround → Friction → Gap → Opportunity chain and returns a citation-backed report you can act on.
Read the docsVisit Qyntel
Free tier · 1 validation / month · then paid credits
qyntel validate
qazimut qyntel validate "AI contract reviewer for freelance designers" --output markdown --mock
✔ Report generated (ID: qnt_994821)
Verdict: Worth Validating Further
command reference
Every command, documented.
Synopsis, examples, flags, presets and output samples for all eight Qazimut commands. Search to jump straight to a flag or example.
qazimut chat
qazimut chat
Interactive AI chat that ties the whole utility belt together.
Synopsis
qazimut chat ["prompt"] [flags]
Examples
Open the interactive TUI
qazimut chat
Compress images via natural language
qazimut chat "compress the images in ./public"
Explain piped input
npm test 2>&1 | qazimut chat "explain this"
Switch provider on the fly
qazimut chat --provider ollama --model llama3.2
Flags
FlagDescription
--provider <name>openai | anthropic | google | ollama | lmstudio | llamacpp | openrouter
--model <name>Provider-specific model name
--jsonMachine-readable output
--no-colorDisable ANSI colors
--verboseVerbose logging
--config <path>Custom config path
Output example
  ██████╗  █████╗ ███████╗██╗███╗   ███╗██╗   ██╗████████╗
  ██╔═══██╗██╔══██╗╚══███╔╝██║████╗ ████║██║   ██║╚══██╔══╝
  ██║   ██║███████║  ███╔╝ ██║██╔████╔██║██║   ██║   ██║
  ██║▄▄ ██║██╔══██║ ███╔╝  ██║██║╚██╔╝██║██║   ██║   ██║
  ╚██████╔╝██║  ██║███████╗██║██║ ╚═╝ ██║╚██████╔╝   ██║
   ╚══▀▀═╝ ╚═╝  ╚═╝╚══════╝╚═╝╚═╝     ╚═╝ ╚═════╝    ╚═╝

  v0.3.1 · ollama:llama3.2 · cwd-aware
  ────────────────────────────────────────────────

  > compress the images in ./public
  ⏺ invoking img compress --preset web ./public …
  ✔ compressed 12 files (18.2 MB → 1.1 MB · -94.0%)
Related commands
qazimut imgqazimut learnqazimut explain
faq
Questions, answered.
The short version of everything we get asked. The long version lives in the docs.
The Qazimut core (chat, learn, img, db, commit, explain, config, setup) is MIT-licensed and free. You bring your own provider key, or run a local model like Ollama for zero cost. The qyntel subcommand calls Qnaku's paid Qyntel engine — there's a free tier (1 validation/month) and then paid credits.