Terminal-native AI companion

ifq Bring AI back to the terminal.

Less switching. Fewer interruptions. Zero waste.
A question comes up. Just ask. The answer's right by your cursor.

Install with npm (opens in a new tab)GitHub (opens in a new tab)
Quick start

Click to copy

ifq — zsh
$ ifq
ifq v1.0 · gpt-4o · session: default
ifq > explain why curl returns 403
403 means the server refused the request. Common causes: a missing auth header, IP restrictions, or a missing CSRF token…
ifq > /deep Redis vs Memcached
⟳ querying 4 models in parallel…
Expert consensus: Redis fits persistence + rich data structures; Memcached fits pure caching and dead-simple setup…
0

Zero dependencies

Pure Node.js · feather-light

50+

Commands & directives

Covers your whole AI workflow

6+

AI providers

OpenAI · Claude · DeepSeek · Ollama…

∞

Persistent memory

Context that never drops

A new default

A question comes up.

Don't switch windows.

Let ifq answer.

Core capabilities

Not just chat. A complete workflow.

Nine dimensions. Everything you need from AI in the terminal.

⚡Immediate

Instant answers

ifq "your question" — one line, without leaving the terminal. Ask, translate, look up, write, review — every step gets shorter.

ifq askifq translateifq shell
🧠Deep Analysis

Deep analysis

Query several models in parallel, then synthesize an expert consensus. For the big calls, get more than one opinion.

ifq deep
💬Persistent

Persistent sessions

Context never drops. Automatic semantic compaction. Isolated sessions — one each for work, ops, and learning.

ifq chatifq --session work
🧭Project Context

Project setup

Generate .ifq, IFQ.md, and a code index in the current directory. Get project context in place before you ask or review.

ifq initifq configifq --lang zh
🔍Code Review

Code workflow

Review a git diff directly. Generate a Conventional Commit in one shot. Scan across security, performance, and quality.

ifq reviewifq commit
▦Automation

Scriptable output

Built for scripts, CI, and automation. Quiet, no-color, JSON — drops right into your own toolchain.

ifq --jsonifq --quietifq --no-color
🤖Agent Team

Multi-agent teamwork

Auto-detect available AI tools, then classify, route, and synthesize across agents. From solo to in sync.

ifq teamifq route
🧬Agent OS

Self-evolving

Versioned prompts. Auto-evolve system / agent / soul. Diff, roll back, encrypted storage.

ifq evolveifq skills
🎨Design

IFQ Design Skills

12 pro modes (launch film / keynote / dashboard / whitepaper / business card / brand system…) + 24 hand-drawn SVG icons. One line back a deliverable design, signed by ifq.ai by default.

ifq design initifq design modesifq design new M-08
How it works

Four steps. From install to evolve.

01Connectifq config --url https://api.openai.com

Set your API. Auto-detect the provider and available models.

02Chatifq

Drop into a persistent session. Context holds automatically.

03Go deep/deep compare microservice frameworks

Parallel analysis across models. Expert consensus in one step.

04Evolve/evolve

Self-tuning prompts. The more you use it, the better it knows you.

Today

It's already good today.

  • Direct Q&A: ifq ask or ifq "question"
  • Continuous chat: context holds. Memory compacts automatically.
  • Command-line translation: ifq explain · ifq shell
  • Code review: ifq review · ifq commit
  • Deep analysis: ifq deep — multi-model consensus
  • Self-evolution: ifq evolve — versioned prompts
  • Skill system: ifq skills — install, create, invoke
Providers

Connect the models you love.

  • OpenAI
  • Anthropic
  • OpenRouter
  • DeepSeek
  • Ollama
  • Any OpenAI-compatible API

Auto-detect providers. Switch models with one command. ifq m for interactive picking.

$ ifq m
1. gpt-4o ✓
2. claude-sonnet-4-20250514
3. deepseek-chat
Select model [1-3]:
System architecture

Five layers. Each one amplifies.

›Entry

One line in. Zero friction.

⎋Routing

Classify the question. Dispatch it smartly.

⚙Execution

One model for speed. Many for depth.

◉Memory

Semantic compaction. Persistent storage.

∞Evolution

Prompts that iterate on themselves. Stronger the more you use it.

What comes next

Every step, a step further.

v1.1Runtime Integration

Deeper shell integration. Native pipe support. Real-time workflows.

v1.2Production Ready

Stable API. Performance tuning. Error recovery. Production-grade reliability.

v1.3Intelligence Layer

Learns your patterns. Predicts your intent. Suggests proactively.

v1.4Ecosystem

A skills marketplace. A plugin system. From tool to platform.

Try these

Click to copy. Paste and run.

FAQ

Frequently asked questions

What is ifq CLI?

ifq CLI is a command-line AI tool by IFQ.AI (created by peixl) that brings AI directly into your terminal workflow — ask questions, explain code, review diffs, generate shell commands, run multi-model deep analysis, and more, without ever leaving the shell.

How do I install ifq CLI?

Install it globally with npm: npm install -g @peixl/ifq. Then set your API key with ifq config --key sk-your-api-key and run ifq to start a session.

Which AI models and providers does ifq CLI support?

ifq CLI works with OpenAI, Anthropic (Claude), OpenRouter, DeepSeek, Ollama, and any OpenAI-compatible API. It auto-detects providers and lets you switch models with a single command (ifq m).

Is ifq CLI free and open source?

Yes. ifq CLI is free to use and published on npm under the MIT license as @peixl/ifq.

What can ifq CLI do?

Instant Q&A (ifq ask), code explanation (ifq explain), code review (ifq review), commit messages (ifq commit), natural-language shell commands (ifq shell), multi-model deep analysis (ifq deep), persistent sessions (ifq chat), multi-agent teamwork (ifq team), self-evolving prompts (ifq evolve), and IFQ Design Skills (ifq design). Over 50 commands and directives in total.

What is IFQ Design Skills?

IFQ Design Skills is a design system built into ifq CLI: 12 pro modes (launch film, keynote, dashboard, whitepaper, business card, brand system and more) plus 24 hand-drawn SVG icons. One line gives you back a shippable design, signed by ifq.ai by default. See https://cli.ifq.ai/design.

Can ifq CLI run offline or with local models?

Yes. With Ollama — or any local OpenAI-compatible endpoint — ifq CLI can run fully local models offline, so no data leaves your machine.

What are the system requirements?

ifq CLI runs on macOS, Linux, and Windows. It requires Node.js (version 18 or newer) and ships with zero dependencies (pure Node.js, feather-light).

One more thing

This is just the beginning.

Today, it answers.

Tomorrow, it collaborates.

Beyond that, it'll know what you need before you do.

It will remember you.
It will judge for you.
It will orchestrate models.
It will organize intelligence.

One day you'll open the terminal, and it's already waiting.