MacArchiveAll appsMacAItermMacCADNotePadXmacPaint
MacAIterm

A macOS terminal with a built-in agent

MacAIterm combines a real shell with an AI agent that works in the same session. It adds resource monitoring, file trees and saved sessions for work on remote servers over SSH.

v1.0.0macOS 13+Apple SiliconFree · MIT license
Recorded from the running app.
A local model (Ollama) finds and fixes a failing test.
Features

Built around a real shell

MacAIterm does not replace your terminal. The agent, the approval flow and the server tools are added around a real zsh session that you can still type into.

01

An agent in your shell

The agent runs commands in the same session and directory you are working in. It reads, edits and creates files, and shows every change as a diff that can be undone. When a test fails, it continues until the test passes or it stops and explains why.

02

Approvals

Risky commands wait for your approval. The app provides four permission modes, secret masking, a network-block mode and an audit log of every command the agent ran or attempted.

03

Local or cloud models

Use a local model through Ollama, an OpenAI-compatible server or the OpenAI API, or Codex CLI with a ChatGPT login. Local models keep the conversation on your Mac. The other options send it to the provider you select.

04

Remote work over SSH

After you connect with SSH, the agent can run commands, edit files and search on the server. Anything that changes the server requires an approval that names it. Read-only checks run immediately.

05

Resource monitoring

A strip under the tabs shows the server's CPU, memory, GPU (including NVIDIA) and disk. Your Mac's CPU, GPU and memory remain in the top bar. Only a read-only script runs on the server.

06

Sessions and file transfer

Saved sessions, ~/.ssh/config import, jump hosts, X11 and port forwarding. File trees for the server and your Mac open side by side, and dragging copies files with SHA-256 verification. Passwords are never stored.

Screenshots

The application in use

Every image is a capture of the running application. The agent in these screens is Codex CLI, and the SSH screens show a real NVIDIA DGX Spark.

Plan mode

Review the plan before execution

For a request that takes several steps, the agent first proposes a plan of three to six steps. You can approve it, request changes or cancel.

Plan mode: the agent lists the steps it intends to take and waits for approval
Approvals

Approval before risky commands

Deleting, installing, pushing and any other change to the system requires approval, shown with the exact command and the reason. Allow it once, for the session or for the project, or deny it and give the agent feedback.

An approval bar for deleting a file, with once, session, project and deny choices
Local model

Local models for lightweight lookups

An Ollama model runs entirely on your machine, which suits short lookups like this one. Multi-step work needs a Mac with more memory than a 16 GB laptop provides.

A local model answers a simple question by running a lookup command
SSH

Server resources and files in one window

A strip under the tabs shows the server's CPU, memory, GPU and disk, while the top bar keeps your Mac's figures. File trees for both machines open on the left.

An SSH session to a DGX Spark with the server's CPU, memory, GPU and disk under the tabs and file trees on the left
Transfers list showing completed, verified copies

Verified file transfer

Copy files and whole folders in either direction. Each file is verified with SHA-256 before it replaces anything, and a cancelled transfer never leaves a partially written file.

Server status panel with graphs next to this Mac

Detailed server status

The status panel graphs recent samples and compares the server with your Mac. It reads /proc, df and nvidia-smi, and installs or writes nothing.

Agent on the server

The agent works on the remote host

Commands run in the terminal you are watching, in the server's shell. The result card names the server and its folder, so the location of every command is clear.

The agent runs nvidia-smi on the server and reports utilization and temperature
Server approvals

Approvals name the target server

Read-only checks run immediately. Anything that could change the server asks first, and the approval shows the server's name. “Always allow” rules apply only to that server and only for the session.

An approval for a delete command on the server, showing the server's name
Remote GUI

Remote GUI over SSH

Shows a server's VNC desktop inside the app, from any SSH tab. Screen data is carried only by that tab's SSH connection, and no port is opened on your Mac. The VNC password is requested when you connect and is never stored. The AI chat does not view or operate the screen. Requests run as shell commands over SSH and the results appear in the chat, so “open a terminal window” runs a command instead of clicking.

Remote GUI: the XFCE desktop of an Ubuntu server with a file manager inside the window, and the AI chat on the right running whoami as a shell command
SSH config

Edit ~/.ssh/config in the app

Opens ~/.ssh/config in a window, from the session panel. The syntax is checked with the real ssh -G before saving, and errors are reported with their line number. Saving updates the host list at once and keeps the previous content as a timestamped backup. A file that changed elsewhere in the meantime is never overwritten.

The ~/.ssh/config editor: the file as text with the selected host, a valid-syntax message, and the Check syntax and Save and apply buttons
Session manager with saved sessions, folders and hosts from ~/.ssh/config

Session manager

Saved sessions in folders, hosts from ~/.ssh/config and recent logins. Double-click to connect. Passwords stay in the terminal, where you type them.

Session editor: key file, jump host, X11 forwarding and port forwards

Session settings

Key file, jump host, start folder, X11, port forwarding and keep-alive. Every field is validated, so a pasted value cannot become an additional ssh option.

Models

Model providers

No model is bundled. Choose where the agent runs, and change it at any time from the chip in the top bar.

Option Requirements Notes
Local (Ollama) Ollama and a model, for example ollama pull qwen3:8b The conversation stays on your Mac.
An 8B model fits a 16 GB Mac but is slow for multi-step work.
OpenAI API / compatible An OPENAI_API_KEY, or the address of a server such as LM Studio, llama.cpp or vLLM The context is sent to that server.
Keys are read from the environment or stored in the Keychain and are never displayed.
Codex CLI codex signed in with ChatGPT The context is sent to OpenAI through your Codex login.
The screenshots above were captured this way.
Getting started

Set up in three steps

01

Install

Download the .dmg, drag MacAIterm to Applications and, on the first launch, right-click and choose Open. The app is not signed yet.

02

Choose a model

Pull a model with Ollama, add an API key or sign in to Codex. Then select it in the model chip.

03

Send a request

For example, “Which process is using port 3000?” or “This test fails; fix it.” For a server, connect with ssh as usual.

Download MacAIterm

Free, MIT licensed, no account. Version 1.0.0 for Macs with Apple silicon.

Go to the download page
Version
1.0.0
Requires
macOS 13 or later, Apple silicon
Size
130 MB
Signed
No.
The first launch needs right-click → Open