terminaI

Name: terminaI
Author: Prof-Harita

Issues

Open-source, local-first alternative to Cowork-style computer assistants: real PTY terminal ops, explicit approvals, JSONL audit logs. Windows + Linux + macOS. Model agnostic.

418stars

37forks

TypeScript

Installation

# Add to your Claude Code skills
git clone https://github.com/Prof-Harita/terminaI

Getting Started

Guides for using ai agents skills like terminaI.

Caveman: Cut Claude Token Use by 65%
How agent-side prompt compression works, when to use it, and when not to.
What is an AI Skills Marketplace?
Definitions, how marketplaces work, and how to choose between them in 2026.
Getting Started with AI Skills

Security ReportIssues

Last scanned: 5/22/2026

{
  "issues": [
    {
      "type": "npm-audit",
      "message": "@azure/identity: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "@azure/msal-node: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "@google-cloud/storage: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "@isaacs/brace-expansion: @isaacs/brace-expansion has Uncontrolled Resource Consumption",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "@modelcontextprotocol/sdk: Anthropic's MCP TypeScript SDK has a ReDoS vulnerability",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "@opentelemetry/exporter-prometheus: Prometheus exporter process crash via malformed HTTP request",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "@opentelemetry/sdk-node: Vulnerability found",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "@protobufjs/utf8: protobufjs has overlong UTF-8 decoding",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "@tootallnate/once: @tootallnate/once vulnerable to Incorrect Control Flow Scoping",
      "severity": "low"
    },
    {
      "type": "npm-audit",
      "message": "ajv: ajv has ReDoS when using `$data` option",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "body-parser: Vulnerability found",
      "severity": "low"
    },
    {
      "type": "npm-audit",
      "message": "brace-expansion: brace-expansion: Zero-step sequence causes process hang and memory exhaustion",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "diff: jsdiff has a Denial of Service vulnerability in parsePatch and applyPatch",
      "severity": "low"
    },
    {
      "type": "npm-audit",
      "message": "esbuild: esbuild enables any website to send any requests to the development server and read the response",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "express: Vulnerability found",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "fast-uri: fast-uri vulnerable to path traversal via percent-encoded dot segments",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "fast-xml-parser: fast-xml-parser has an entity encoding bypass via regex injection in DOCTYPE entity names",
      "severity": "critical"
    },
    {
      "type": "npm-audit",
      "message": "flatted: flatted vulnerable to unbounded recursion DoS in parse() revive phase",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "gaxios: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "jws: auth0/node-jws Improperly Verifies HMAC Signature",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "lodash: Lodash has Prototype Pollution Vulnerability in `_.unset` and `_.omit` functions",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "markdown-it: markdown-it is has a Regular Expression Denial of Service (ReDoS)",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "minimatch: minimatch has a ReDoS via repeated wildcards with non-matching literal in pattern",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "path-to-regexp: path-to-regexp vulnerable to Regular Expression Denial of Service via multiple route parameters",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "picomatch: Picomatch: Method Injection in POSIX Character Classes causes incorrect Glob Matching",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "postcss: PostCSS has XSS via Unescaped </style> in its CSS Stringify Output",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "protobufjs: Arbitrary code execution in protobufjs",
      "severity": "critical"
    },
    {
      "type": "npm-audit",
      "message": "qs: qs's arrayLimit bypass in comma parsing allows denial of service",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "retry-request: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "rollup: Rollup 4 has Arbitrary File Write via Path Traversal",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "simple-git: simple-git Affected by Command Execution via Option-Parsing Bypass",
      "severity": "critical"
    },
    {
      "type": "npm-audit",
      "message": "tar: node-tar Vulnerable to Arbitrary File Creation/Overwrite via Hardlink Path Traversal",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "teeny-request: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "turbo: Trubo: Login callback CSRF/session fixation",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "underscore: Underscore has unlimited recursion in _.flatten and _.isEqual, potential for DoS attack",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "undici: Undici has an unbounded decompression chain in HTTP responses on Node.js Fetch API via Content-Encoding leads to resource exhaustion",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "uuid: uuid: Missing buffer bounds check in v3/v5/v6 when buf is provided",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "vite: Vite Vulnerable to Path Traversal in Optimized Deps `.map` Handling",
      "severity": "high"
    },
    {
      "type": "npm-audit",
      "message": "vite-node: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "vitest: Vulnerability found",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "ws: ws: Uninitialized memory disclosure",
      "severity": "medium"
    },
    {
      "type": "npm-audit",
      "message": "yaml: yaml is vulnerable to Stack Overflow via deeply nested YAML collections",
      "severity": "medium"
    }
  ],
  "status": "FAILED",
  "scannedAt": "2026-05-22T07:44:29.733Z",
  "semgrepRan": false,
  "npmAuditRan": true,
  "pipAuditRan": true
}

README.md

Frequently Asked Questions

What is terminaI?

terminaI is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by Prof-Harita. Open-source, local-first alternative to Cowork-style computer assistants: real PTY terminal ops, explicit approvals, JSONL audit logs. Windows + Linux + macOS. Model agnostic. It has 418 GitHub stars.

Is terminaI safe to use?

terminaI failed SkillsLLM's automated security scan, which flagged one or more high-severity issues. Review the Security Report section carefully before using it.

How do I install terminaI?

Clone the repository with "git clone https://github.com/Prof-Harita/terminaI" and add it to your Claude Code skills directory (see the Installation section above).

What programming language is terminaI written in?

terminaI is primarily written in TypeScript. It is open-source under Prof-Harita on GitHub, so you can review or fork the full source.

Are there alternatives to terminaI?

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh terminaI against similar tools.

Agentic AI for Beginners

Build your first AI agent from scratch - tool use, ReAct pattern, memory, deployment

41 minBeginner

Comments (0)

to leave a comment.

No comments yet. Be the first to share your thoughts!

Related Skills

superpowers

by obra

An agentic skills framework & software development methodology that works.

234,966

entroly agent-resources

TerminAI

The safe, local-first AI operator for your terminal.

Stop copy-pasting from ChatGPT. TerminAI is the CLI that operates your machine with permissions you control.

Runs on Windows/Linux/macOS using the models you already trust (Gemini, OpenAI, Local LLMs). Designed for real system changes or user workflows - governed, reviewable, and reversible.

TerminAI is an AI-powered system operator.
You describe what you want; TerminAI inspects your machine, proposes a plan, and executes it through governed tools with a deterministic approval ladder and an audit trail.

Status: Public Preview (v0.x)
The core operator loop is rock-solid. CLI is the canonical surface. Desktop/Voice/A2A are optional and still being hardened across platforms. Expect power; expect rough edges. Contributions welcome.

What this is / what this is not

This is:

A local-first operator runtime: intent → governed plan → approvals → execution → audit log.
A real terminal operator (true PTY): it can handle interactive sessions (sudo prompts, ssh, vim/htop-style flows).
A model-flexible setup: choose a provider you already use (including OpenAI-compatible endpoints and local gateways).
100% free and 100% private if you want it to be. Use free models on OpenRouter, run local LLMs—no telemetry, no charges, ever.

This is not:

A hosted service. There is no "contact sales," no pilots, no managed deployment, and no support obligations—this is best-effort OSS.
An "unchecked agent" that silently runs commands. Risky actions are gated behind explicit approvals and logged locally.
A promise that every surface is production-hardened today (desktop/voice/remote features are optional previews).

Demo

Try it right now (10 min)

Pick one of these prompts and paste it into TerminAI. (Click to copy the full prompt)

"Plan an itinerary for Paris. I am visiting as a tourist for 2 days next Saturday. I have already seen the big attractions (Eiffel, Louvre). Find me hidden gems and plan my trip hour-by-hour, including restaurants. Show it on a map. I will visit all these locations via the Metro. Once done, save the map-based visual output as a PDF."

"My graphics drivers seem unstable after the last OS update. Identify my GPU model, check the currently installed driver version, and find the correct latest stable driver. Propose a clean install plan and verify it before execution."

"Check the price of Bitcoin every 10 minutes. If it drops below $90,000, send a system notification and append the timestamp/price to ~/crypto_log.csv. Keep running this in the background until I stop it."

Vision

What if your computer could do things for you safely?

Not just suggesting commands. Not just generating scripts. Actually operating your system—diagnosing issues, fixing problems, and orchestrating workflows—with guardrails that make execution trustworthy.

For years, "internet help" meant: copy instructions from a chat window, paste into your terminal, and hope you didn't miss a step. Coding tools (Cursor, Aider, etc.) reduced that friction for software development—but the rest of your computer is still stuck in the copy/paste era.

TerminAI is the next interface: you state intent, it proposes a governed plan, you approve at the right level, and it executes with an audit trail. Fewer fragile steps. More outcomes—without handing your machine to an unchecked agent.

Why this exists

Computers are powerful, but operating them still requires memorizing syntax, hunting through logs, and repeating the same diagnosis steps.

TerminAI compresses the gap between intent and execution—without turning your machine into an uncontrolled "agent."

Intent → Governed plan → Approval → Execution → Audit

Who it's for

Generalist users — “Review the largest files in my downloads folder, prioritizing the oldest first, i need to free up 200 GB of space.” “What is wrong with my internet; reset my adapter.” “Setup a HP printer.” "Monitor the stock price of NVDA and alert me when it hits xx$". "Why is my system so slow? Can you free up memory" "Can you install the best markdown viewer on my linux machine?". "I just re-installed Windows. Find specific drivers for my GPU and install them".
Developers — “Set up my machine for this repo.” “Fix my dev environment.” “Explain what broke and how to recover.” "Install all the dependencies for this project"
Tech purists — AI leverage with deterministic guardrails, explicit approvals, and tamper-evident logs.
Professional engineers / DevOps — Remotely control your servers, repeatable workflows, headless-friendly CLI, safer execution, readable audit trails.

Two ways to use TerminAI

CLI (canonical): for developers, power users, and headless environments.
Desktop (preview): a Tauri wrapper around the same engine (GUI + voice surface). Linux/Windows are early; macOS is in progress. Treat as experimental until hardened.

The canonical command is: terminai.

What makes TerminAI different

AI system operation must be governed. TerminAI is built around that primitive.

True PTY support: handles interactive applications correctly (sudo prompts, ssh sessions, vim/htop-like flows) via node-pty.
Governed autonomy: a strict policy engine + deterministic approval ladder (A/B/C) replaces "yolo execution."
Audit trail by default: structured JSONL logs for actions and intent tracking (local files you can review/replay).
Provider flexibility: choose what you already use—Google, OpenAI sign-in, or any OpenAI-compatible /chat/completions endpoint (OpenRouter, local gateways, etc.).
Deep extensibility: first-class support for MCP (tool connectors) and an optional Agent-to-Agent (A2A) control plane (preview).
Local-first posture: core operation runs on your machine; no hosted relay is required. Any optional remote features are opt-in and under your control.

Install

CLI (npm)

Prereqs: Node.js 20+

npm i -g @terminai/cli
terminai --version
terminai

If you don't see the CLI package yet (or want bleeding-edge), install from source below.

Desktop (installers) — preview

Download installers from GitHub Releases (desktop is in preview):

https://github.com/Prof-Harita/terminaI/releases

Git clone and build (contributors / bleeding-edge)

git clone https://github.com/Prof-Harita/terminaI.git
cd terminaI
npm install
npm run build

Upgrade (if you installed earlier)

If you installed before the recent multi-provider updates, upgrade so you get the latest auth + platform parity.

npm global:

npm i -g @terminai/cli@latest
terminai --version

release installers: grab the latest from Releases and reinstall:
- https://github.com/Prof-Harita/terminaI/releases
source build:
```
git pull
npm install
npm run build
```

Quick start (providers)

Run terminai and the wizard guides you through setup.

Google Gemini (default)

Fastest path for many users.

terminai
# Select "Google Gemini" → browser opens → sign in

Or use an API key:

# macOS/Linux
export TERMINAI_API_KEY="your-gemini-key"

# Windows (PowerShell)
$env:TERMINAI_API_KEY='your-gemini-key'

terminai

ChatGPT sign-in (OAuth) — preview

Use an OpenAI browser sign-in flow (similar to how Codex tooling supports "Sign in with ChatGPT").

terminai
# Select "ChatGPT Sign-in (OAuth)" → sign in with OpenAI

Note: this is still being hardened. If anything fails, attach logs (see below) and open an issue.

OpenAI-Compatible

Connect to OpenAI Platform, OpenRouter, Ollama gateways, or any /chat/completions endpoint.

# macOS/Linux
export OPENAI_API_KEY="sk-..."

# Windows (PowerShell)
$env:OPENAI_API_KEY='sk-...'

terminai
# Select "OpenAI Compatible" → enter base URL and model

Popular configurations:

OpenAI: https://api.openai.com/v1
OpenRouter: https://openrouter.ai/api/v1
Local LLM gateway: http://localhost:11434/v1

See: docs-terminai/multi-llm-support.md

Use `/llm reset` to switch when inside. Note right now hot-swap is not fully
live. Please restart (Ctrl-C and npm run start or terminai) to apply changes.

The Sovereign Runtime (Safety & Architecture)

TerminAI operates on a 3-Layer Defense-in-Depth model to ensure that autonomous execution never becomes "unchecked access":

Cognitive Layer (The "Brain"): A PAC Loop (Plan-Act-Check) with multi-model consensus that verifies intent before proposing action.
Governance Layer (The "Guardrails"): A deterministic Policy Engine with an Approval Ladder (A/B/C) that forces explicit human sign-off for mutating or destructive actions.