AgentUI for AI Agents

Build with the agent of your choice.

Any AI works with AgentUI. Install our specialized CLI, log in once, and let your agent handle the rest — sync workflows, edit components, push changes, deploy.

Install the CLI
~/your-projectagentui
# 1. install
$ npm install -g @agentuiai/cli
# 2. login (OTP via email)
$ agentui auth login
→ Verification code sent to you@team.com
# 3. hand off to your AI
$ claude "sync workflow_42 and add audit trails"
running— agentui project sync 42 --json
12 components synced
$
01 / 03

Bring your own AI

Claude Code, Cursor, Aider, Codex, your own custom agent. AgentUI doesn't care which AI assistant you run locally — the CLI gives them all the same clean interface to read, edit, and ship AgentUI projects.

02 / 03

Specialized for AI

Every command supports --json. Deterministic exit codes. Machine-parseable output by default. The same approach as the AWS and GitHub CLIs — built for the way agents read tools.

03 / 03

Install. Login. Done.

Three commands and your agent is in. No SDK to wire, no API surface to learn — point any AI at the CLI and it can read, write, and ship AgentUI projects.

From terminal to production in three commands

1

Install

$ npm install -g @agentuiai/cli

One global install. Node 22+. Verify with agentui --version.

2

Login

$ agentui auth login

Email OTP, six digits. Same login as the AgentUI web app — no token juggling.

3

Hand it to the agent

$ agentui project sync <id>

Your agent reads, edits, and pushes from the local working tree. Git, VS Code, Cursor, Claude Code — pick one, the CLI gets out of the way.

Tested with

Claude Code, Codex, Aider, Cursor agent mode, Continue, Cline, custom agents you wrote last weekend — they all read --json. The CLI gives every agent the same surface, so the model you pick doesn't matter to AgentUI.

Claude CodeCodexAiderCursor agentContinueClineRoo CodeOpenAI o-seriesGemini CLILocal llama.cpp…and yours
Made for machines

Output your agent can actually parse

01

--json on every command, schema-stable across versions

02

Deterministic exit codes — agents can branch on success vs validation vs auth failure

03

Dry-run on every mutating command, so the agent can plan before it commits

04

Multi-tenant safety: workspace selection is explicit, never implicit

The secure vibe coding platform

Build fast. Ship safe.

Vibe coding is fast — but most platforms hand you a prototype and call it production. AgentUI hardens, scans, and gates every app you build, so what you ship is what your team can actually trust.

SOC 2 AI app

AES-256 encryption, exportable audit logs, RBAC mapped to Trust Services criteria. Pass your SOC 2 audit on the AI part.

Secure AI hosting

Tenant isolation, encrypted secrets, continuous scanning. Hosting that defends — not just configures.

Enterprise vibe coding

Vibe coding for the enterprise — without the compliance debt. Audit trail, RBAC, security review baked in.

Open a terminal. Let the agent build.

Free to install. Same login as your AgentUI workspace. Pair it with the AI of your choice.

npm install -g @agentuiai/cli
Install the CLIBack to home

Node 22+ · MIT licensed · Same OTP as your AgentUI workspace