FIG 4.12

Your whole AI stack, one command

The Hiveflow CLI routes Claude Code, Codex, Gemini and Aider, ships its own agent with a full tool-use loop, and runs local workers off your kanban boards. MIT, auditable, on macOS, Linux and Windows.

One terminal entry point with three layers. The router reads the task and launches the coding CLI that is best at it. The native agent does the work itself when you would rather not depend on any external CLI. And workers watch your Hiveflow boards, turning cards into pull requests with human gates where you want them.

1

Install it

One line, no npm or git required. MIT-licensed, so you can read every line first

curl -fsSL https://hiveflow.ai/install.sh | bash

or with npm

npm install -g @hiveflow/cli
2

Sign in

Run hiveflow and authorize with your account. Shortcuts: hive · hf · flow

hiveflow
3

Route, or run the native agent

It picks the best installed CLI for the task, or runs its own tool-use loop with /agent

hiveflow ❯ fix the login bug hiveflow ❯ /agent refactor src/auth.js and run the tests hiveflow ❯ /worker list
Using Claude Code or Codex? The Hiveflow CLI is the better door.GitHub

Task-type router

It infers whether the job is a quick fix, a refactor or research, and hands it to the installed CLI that fits.

Native agent

Its own tool-use loop: reads, edits, runs commands, searches and delegates to subagents, with permissions and sessions.

Workers on your boards

Local agents watch kanban columns and work the cards, in parallel, with human-in-the-loop gates and metrics.

Skills, hooks and MCP

The same MCP connector as every other client, plus skills and hooks that load themselves when relevant.

Running in one minute

01

Install

curl -fsSL https://hiveflow.ai/install.sh | bash — or, if you prefer npm, npm install -g @hiveflow/cli

02

Sign in

Run hiveflow and authorize with your account. Shortcuts: hive, hf, flow.

03

Ask for anything

Type the task and let it route, or use /agent and /worker for the other two layers.

Build it on Hiveflow today