Skip to content
v2.2.0 · Free to start

The AI coding agent in your terminal — one key for every model.

Sign in, create a Vinus key, and pick a plan with generous included usage — no per-token math. GPT, Claude, Gemini and more through one hosted Vinus endpoint, straight from the CLI.

$npm install -g vinuscli
21+
bundled skills
5
built-in connectors
1
key, every model
3
operating modes

Every frontier model. One key.

One Vinus key, with an automatic fallback chain.

GPTClaudeGeminiLlamaMistralDeepSeekQwenGrokGPTClaudeGeminiLlamaMistralDeepSeekQwenGrok

Features

Not a chatbot. An operator.

Everything an engineer does in a shell, the agent does too — with your approval, at the level you choose.

Any model, one key

GPT, Claude, Gemini, Llama, Mistral… every frontier model through one hosted Vinus endpoint, with an automatic fallback chain when one is down.

GPTClaudeGeminiLlamaMistralDeepSeekQwenGrok

Real tools, real work

Reads and writes files, runs shell commands, searches the web and GitHub, fetches URLs. Work happens in your repo, not in a sandbox.

Diffs before disk

Every change is previewed as a green/red visual diff before it touches your files. You stay in control, Claude-Code style.

Self-improving skills

Progressive-disclosure knowledge packs. 21 bundled — and the agent writes its own when something is worth remembering.

Your Vinus key

A Vinus account issues a vk- key for our OpenAI-compatible API — usage billed as fair-use credits, never per token. Paste it into the CLI with /login and you're connected.

Reverse engineering

First-class IDA Pro / Ghidra workflow: decompile, cross-reference, patch — straight from the conversation.

Demo

Watch it work.

A real VinusCLI session — it plans, edits with a diff, runs the build, and ships.

vinus — live sessionrec

Operating modes

Autonomy is a dial, not a switch.

Switch at any time — the current mode is always visible in the prompt.

1

Plan

/plan

Read-only. Investigates your codebase and returns a numbered plan — never writes a file, never runs a command.

AutonomyRead-only
2

Manual

/manual

Acts, but asks. Every write, delete and shell command needs your explicit approval before it happens.

AutonomyAsk every time
3

Auto

/auto

Full autonomy. Ships end-to-end and only pauses when genuinely unsure about a destructive step.

AutonomyFull autonomy

Skills

An agent that remembers what it learns.

Skills are progressive-disclosure knowledge packs: a name and one line in the prompt, the full playbook loaded only when needed.

01

Progressive disclosure

Light in the prompt, deep on demand — skills keep the context window lean.

02

Self-authoring

When the agent solves something hard, it writes the skill so it never has to re-derive it.

03

One command away

/skills lists them, /skills install all pulls the official set on top of the 21 bundled ones.

--- SKILL.md
name: reverse-engineering
description: IDA Pro / Ghidra — decompile, xrefs, patch
---
skill-creatorfrontend-designmcp-builderreverse-engineeringweb-artifacts-buildertheme-factorycanvas-designwebapp-testingclaude-apiidapythonyour-next-skill…skill-creatorfrontend-designmcp-builderreverse-engineeringweb-artifacts-buildertheme-factorycanvas-designwebapp-testingclaude-apiidapythonyour-next-skill…
your-next-skill…idapythonclaude-apiwebapp-testingcanvas-designtheme-factoryweb-artifacts-builderreverse-engineeringmcp-builderfrontend-designskill-creatoryour-next-skill…idapythonclaude-apiwebapp-testingcanvas-designtheme-factoryweb-artifacts-builderreverse-engineeringmcp-builderfrontend-designskill-creator

Connectors

Plugged into your stack.

Connect external services on demand — their tools become the agent's tools.

GitHub

Repos, pull requests, issues.

Canva

Posters, CVs and slides from chat.

Notion

Pages and databases, agent-ready.

Linear

Issues and projects from the terminal.

IDA Pro

A loaded binary becomes a conversation.

MCP

Any MCP

Bring your own server with /mcp.

OAuth happens in your browser. Tokens stay on your machine.

Pricing

One key. Every model. Your plan.

Start free. Upgrade for generous included usage on capable models — fair-use credits, never per-token billing.

Free

The full agent, on your machine.

€0forever
Install now
  • Full agent — all tools, modes, skills, connectors
  • Hosted free models via your Vinus key
  • 40 messages per month included
  • Works in any OpenAI-compatible tool
  • No telemetry
Most popular

Pro

For daily driving on capable models.

€20/mo
Get started
  • Everything in Free
  • Generous included usage on paid models (Claude Sonnet, GPT-4o, Gemini 2.5 Pro…)
  • Fair-use credits — never billed per token
  • Top up with credit packs anytime
  • Usage dashboard

Pro+

For heavy use and top-tier models.

€100/mo
Get started
  • Everything in Pro
  • Top-tier models (Claude Opus, o1…)
  • Highest included usage
  • Top up with credit packs anytime
  • Priority email support

Compare plans

FreeProPro+
Hosted modelsFree tier+ paid models+ top-tier
Messages / month40GenerousHighest
Credit top-ups
Usage dashboard
Priority support
Included usage runs on fair-use credits. Reach the limit and heavier models pause until reset — or top up with a credit pack. You never pay per token.
Every plan runs on one Vinus key — usage is billed as fair-use credits, never per token.

Install

Thirty seconds to your first agent.

One line. Requires Node.js ≥ 18.

$
Then, inside vinus:/login vk-…/modelBuild me a REST API with tests
Free to start

A free Vinus account includes a hosted free tier — start without paying, upgrade when you need more.

New

Create your Vinus key on the site, run /login in the CLI, and go. Your plan's included usage is billed as fair-use credits, never per token.

Create your key on the dashboard — a free tier is included.

FAQ

Questions, answered.

Everything people ask before their first vinus session.

Yes — the CLI is free, with no telemetry. A free Vinus account includes hosted free models. Paid plans (Pro, Pro+) add generous included usage on more capable models.

$ vinus

Ship from your terminal.

Free to start — with hosted Pro and Pro+ plans when you want more. Fair-use credits, never per-token billing.