Toone
Toone is a native macOS app that turns repeat work described in plain words into routines: a graph of steps run by Claude Code or Codex agents on your own subscription, with per-step contracts, approval gates, memory between runs and pause-edit-resume. Pro from €9.99/month or €199.99 lifetime.
About Toone
What Toone is
Toone is a native macOS workspace for agentic workflows. You describe a multi-step job the way you would brief a colleague (when it runs, which models, which accounts) and Toone drafts it as a routine: a graph of steps that specialist agents carry out, in parallel where the work allows. While it runs, every step lights up on a live graph. A step whose output falls short pauses on its own, you edit it, and the run resumes where it stopped while finished steps keep their results.
The agents are not Toone's own models. Toone runs on your Claude Code or Codex, with the Anthropic or OpenAI subscription you already pay for, and works in your own website accounts through a built-in browser. Toone's role is the layer around those agents: structure, checks, memory between runs and control. Its Product Hunt tagline, "Grokbot meets n8n/zapier", sums up the pitch: a chat-style agent with the repeatability of an automation tool.
Toone is built by Matheus Paranhos, an AI solutions and software engineer, and is published by him as a sole trader in Portugal under his Hexagonal.io brand. The company says it is supported by the Claude Startups and AWS Startups programmes, and it is not affiliated with Anthropic, OpenAI or AWS. Toone placed #3 on Product Hunt on November 18, 2026.
How it works
- Describe a routine. A routine names its inputs, the responsible agents, steps, completion criteria and outputs. Routines can call other routines, and each step can use a different model, for example cheap parallel research on Haiku and a final check on Opus.
- Contracts. Each step's output is checked against a contract you set ("at least 20 leads", "every claim sourced", "brand voice at least 8/10"). Failures follow a route you choose: retry, switch model, or ask you.
- Memory between runs. A routine remembers what it already did, so a weekly lead search does not return last week's leads again. Toone contrasts this with chat assistants that can only schedule a prompt that starts fresh each time.
- Approval gates and receipts. Routines can stop for your approval before sensitive steps, such as publishing. Project History keeps a timeline of agent and human changes, and for Claude agents a Safe Mode screens tool output and records each result in an audit log.
- Mac-native extras. An Agent Spotlight summons any agent over any app; the window resizes from a compact companion to a full workspace; up to nine projects stay open; voice input and meeting capture (microphone plus system audio, with speaker labels) turn meetings into shared context.
- Explore library. A public library of reviewed routines and bundles you can add and adapt. Version 1.0.82 adds automatic steps that run a script on your Mac by themselves and pause the run if they fail.
- Local-first. Agents, routines, run history, conversations and project files live on your Mac; website accounts, analytics, Explore listings and an optional encrypted relay are separate services.
We checked the release feed
Toone's public GitHub repository (io-hexagonal/Toone) holds the website source under MIT and publishes the app's releases. On November 19, 2026 we read the latest release and its Sparkle update feeds:
$ curl -s https://api.github.com/repos/io-hexagonal/Toone/releases?per_page=3
v1.0.82 2026-11-18 Toone-1.0.82.zip (66.6 MB), Toone-Liquid-Glass-1.0.82.zip (66.4 MB), 4 appcasts
v1.0.81 2026-11-16
v1.0.80 2026-10-02
$ curl -sL .../v1.0.82/appcast-standard.xml | grep minimumSystemVersion
<sparkle:minimumSystemVersion>14.0</sparkle:minimumSystemVersion>
$ curl -sL .../v1.0.82/appcast-liquid-glass.xml | grep minimumSystemVersion
<sparkle:minimumSystemVersion>26.0</sparkle:minimumSystemVersion>
Three practical points. There are two builds of the same version: a standard one for macOS 14 Sonoma and later, and a "Liquid Glass" one that needs macOS 26, where Apple introduced that design language. Updates install automatically through Sparkle, so existing installs follow the release feed; two releases shipped in the three days around launch. And the app itself is not open source: the repository's LICENSE-BINARY is a proprietary end-user licence from Hexagonal.io that forbids reverse engineering and redistribution; only the website code is MIT.
Pricing (checked November 19, 2026)
| Plan | Price | What you get |
|---|---|---|
| Free | €0 | Create 2 routines; running routines is never limited. Not available yet: the pricing page says free accounts open in early 2027 |
| Pro Monthly | €9.99/month | Unlimited routine creation, install any routine from Explore |
| Pro Yearly | €79.99/year | Same as Monthly |
| Pro Lifetime | €199.99 once | Same, no renewal |
Prices include VAT where it applies, the first purchase has a 14-day refund, and new web purchases are sold through Link (a Stripe company) as merchant of record. Model usage is billed separately by Anthropic or OpenAI through your own subscription; Toone sells no model plan.
Worked example: which Pro plan to buy
Monthly for a year = 12 x €9.99 = €119.88
Yearly = €79.99 (33% less than 12 months of Monthly)
Lifetime vs Yearly = €199.99 / €79.99 = 2.5 years to break even
Lifetime vs Monthly = €199.99 / €9.99 = about 20 months to break even
Since Pro only lifts the limit on creating routines (runs are unlimited on every plan), the question is how many routines you will build. Someone who needs only three or four stable routines could start with a month of Pro and build them; the pricing page says Pro stays on until the end of the period you paid for and that runs stay unlimited on every plan, but what an account without Pro looks like before free accounts open in 2027 is not spelled out, so check before cancelling. Lifetime only pays off if you expect to keep building new routines for more than two and a half years on a product that is still in early access.
Toone vs the alternatives
| Toone | Nod | Rakazo | n8n / Zapier / Make | |
|---|---|---|---|---|
| Model | Natural-language routines run by agents on a step graph | Personal agent that runs errands | Persistent bots with memory and routines | Node-based automation workflows |
| Where it runs | macOS app (14+) | Mac, Windows or server | Self-hosted or Rakazo Cloud | Cloud; n8n can be self-hosted |
| AI provider | Your Claude Code or Codex subscription | 30+ model providers or a ChatGPT plan | Any model | AI steps via API keys |
| Output checks | Contracts per step, pause and resume | Asks before it acts | Approval rules | Error branches and retries you build |
| Licence | Proprietary app | Apache 2.0 | Apache 2.0 | n8n fair-code; Zapier and Make proprietary |
| Price | Pro €9.99/month or €199.99 lifetime | Free | Free self-hosted; cloud from $15/month | Free tiers to paid plans |
Choose Toone if you want agent-run routines you can watch and correct mid-run on a Mac, and you already pay for Claude or ChatGPT. Choose a classic automation tool if your workflow is deterministic API plumbing that does not need an agent's judgement, or if it has to run on a server around the clock.
Who it's for
Mac users with multi-step work that recurs every week or month: founders, operators, researchers, marketers and creators. The launch examples are a weekly content engine (topic research, drafting, review against a brand-voice contract, visuals in Figma, approval before posting), lead research, and grant and startup-credit applications. Engineering is listed among the starter workflows, but the Explore library on launch day leaned toward founders and marketers.
Limits and trade-offs
- Mac only. There is no Windows, Linux or server version, and routines run while your Mac is on.
- No free plan yet. The FAQ says every account can create 2 routines for free, but the pricing page says free accounts open in early 2027 and the company is starting with paying members.
- You need a Claude or ChatGPT subscription, and routine runs draw down its usage limits.
- Small library. On November 19, 2026 Explore listed six items (bundles and routines), all by the Toone team and all free.
- One-person company. Toone is published by a sole trader; weigh that for business-critical routines, although your routines and data stay on your Mac.
- Proprietary app. Despite the MIT-licensed repository, the app ships under a closed end-user licence.
Related on TechLogHub
- n8n vs Zapier vs Make and AI agents vs AI workflows.
- Claude Code vs Codex CLI vs Gemini CLI, the agents Toone runs on.
- AI agent and prompt engineering.
- Cron expression generator, for the schedules routines run on.
- Boost.space AgentOS and OpenBot, two other ways to run teams of agents.
Value & Audience
Value Proposition
Turn recurring multi-step work into agent routines that run the same steps and checks every time, which you can watch, pause, fix and resume, using the Claude or ChatGPT plan you already have.
Problem Solved
A scheduled chat prompt starts fresh every run, repeats itself and fails silently; recurring agent work needs structure, output checks, memory and human control.
Target Audience
Tech Stack
Social & Links
Loading ratings...
Loading comments...
Gallery

Feedback & Roadmap
Toone FAQ
Common questions about Toone's features, pricing, and use cases
Is Toone free to use?
No. Toone is a paid product, starting at Toone Pro Monthly - €9.99/month - unlimited routine creation, install any Explore routine; runs unlimited (free accounts with 2 routines open early 2027). Plans listed: Toone Pro Monthly - €9.99/month - unlimited routine creation, install any Explore routine; runs unlimited (free accounts with 2 routines open early 2027); Toone Pro Yearly - €79.99/year - same as Monthly; Toone Pro Lifetime - €199.99 one-time - same, no renewal; 14-day refund on first purchase.
What are some alternatives to Toone?
Popular AI Agents & Automation alternatives to Toone include NeverApply, Composio and Harness².
Toone appears in
You May Also Like

NeverApply
AI job-search agent that matches jobs, tailors your resume, and auto-applies.
View ProductComposio
Composio connects AI assistants and custom agents to more than 1,500 apps through managed authentication, SDKs and an MCP gateway, and its new Instant tools let agents pay per call for search, scraping, audio and data services from one balance.
View Product
Harness²
Harness² is an Apache-2.0 research framework from Google Cloud AI Research that automatically refines an LLM agent's harness (prompts, tools, memory, skills and more) for each task and carries lessons forward to future tasks, without changing model weights.
View ProductCraft Agents
Craft Agents is an open-source (Apache 2.0) desktop app from the makers of Craft Docs for working with AI agents. It connects agents to MCP servers, REST APIs, local files and a built-in browser, with multi-session workspaces, Skills, Automations and Pages.
View ProductOne email a week
TechLogHub tracks 903 tools and 611 open-source projects. Get the week's new launches, release roundups and open-source picks — nothing else.
Get the TechLogHub digest
One email a week: new and trending developer tools, fresh comparisons, and what shipped. Unsubscribe in one click.