What each personal agent costs
The sticker price is only half the answer, because agents work for hours and some plans charge that work against your limits. This page lists every plan that includes an agent and explains the catches.
Every plan that includes an agent
| Plan | Price per month | Agent | Details |
|---|---|---|---|
| ChatGPT Pro | $100/mo | Full agent | The entry route to Dots: one Dot, 5x Plus usage in ChatGPT Work and Codex, and unlimited chat with your Dot, which suits most individual professionals. |
| ChatGPT Pro (20x) | $200/mo | Full agent | Still one Dot, but since DevDay 2026 it gives 10x Plus usage instead of 20x and 100 GPT-6 Pro messages a week instead of 200, so check whether the $100 tier already covers you. |
| ChatGPT Pro 500 | $500/mo | Full agent | One Dot with the highest usage limits and Astra Ultrafast at up to 300 tokens per second in Codex and Work, built for people who keep their Dot busy all day. |
| ChatGPT Business Premium | $125/mo per seat or $100 a month on annual billing | Full agent | One Dot for each Premium seat with 5x Standard usage and a two-seat minimum, sold in every ChatGPT region including the EU and UK. |
| ChatGPT Enterprise / Edu / Healthcare | Contact sales per seat | Beta | A Dots beta that an admin has to switch on, with specialist Dots that have their own identity and credentials still in pilots. |
| Muse Free | Free | Limited | Up to 100 million Muse tokens a week at no charge, for adults in the US and Canada who add a payment card. |
| Muse Power | $20/mo | Full agent | 500 million Muse tokens a week with the same features as Maximum, for people who run out of the free allowance. |
| Muse Maximum | $100/mo | Full agent | 3 billion Muse tokens a week, sized for many helper agents that run around the clock. |
| Google AI Pro | $19.99/mo | Limited | The cheapest paid way into Gemini Spark, but only for US subscribers since July 2026, with other countries promised later. |
| Google AI Ultra (5x) | $99.99/mo | Full agent | Gemini Spark with 5x Pro usage in most countries, except the EEA, Switzerland, the UK and Nigeria. |
| Google AI Ultra (20x) | $199.99/mo | Full agent | Gemini Spark with 20x Pro usage plus Google's experimental features, for heavy users who already live in Google apps. |
| Claude Pro | $20/mo or $17 a month on annual billing | Full agent | The merged Claude experience with long tasks, connectors and the Docs and Slides beta, which is enough for most people starting out. |
| Claude Max 5x | $100/mo | Full agent | 5x Pro usage per five-hour session and priority access at busy times, for people who hit Pro limits regularly. |
| Claude Max 20x | $200/mo | Full agent | 20x Pro usage per five-hour session with the same features as Max 5x, for full-time agent work. |
| OpenClaw (self-hosted) | Free | Full agent | Free software with no subscription, where you pay only for the model you connect and the machine that runs it. |
Plans and regions change quickly. Confirm the price and availability on the provider's own page before you pay.
Add up your monthly bill
Choose the agents you want and see the lowest monthly price that covers them.
Totals use list monthly prices in US dollars and ignore usage limits, annual discounts and whether the plan is sold in your country.
What self-hosted OpenClaw really costs
OpenClaw has no subscription and no paid tier. Instead, you pay the model provider for every token the agent reads and writes, either per token through an API key or through a plan you already have. A local model avoids the token bill entirely.
Token prices vary widely. Per million tokens, GPT-6 Luna charges $0.10 for input and $0.50 for output, GPT-6.1 Sol and Claude Sonnet 5.5 charge $2 and $10, Claude Opus 5.5 charges $4 and $20, and GPT-6 Astra charges $10 and $50.
The number to watch is the cached input price. An agent resends its long context on every step, so most of what it reads is repeated. GPT-6.1 Sol bills cached input at $0.10 per million tokens instead of $2, and an agent stuck in a loop can still run up a bill you did not expect.
Run it at home and there is no server bill; rent a small VPS for an always-on agent and expect roughly $5 to $20 a month.
OpenClaw · Claude, GPT, Gemini or local models
Questions
What is the cheapest way to get a personal agent?
Muse Free costs nothing up to 100 million tokens a week, but only in the US and Canada. OpenClaw is free software, so you pay only for the model. The cheapest paid plans are Google AI Pro at $19.99 a month for Gemini Spark in the US and Claude Pro at $20 a month.
Do I need the $500 ChatGPT Pro plan to get Dots?
No. Every ChatGPT Pro tier includes one Dot, starting at $100 a month. The $500 tier only adds the highest limits and Astra Ultrafast speed.
What changed on the $200 ChatGPT Pro plan?
At DevDay 2026 OpenAI reduced its Work and Codex allowance from 20x to 10x Plus usage and cut GPT-6 Pro messages from 200 to 100 a week. Existing subscribers keep their old limits for a transition period and receive a one-time credit.
Does talking to my Dot use up my ChatGPT limits?
Chatting with your Dot does not. Tasks it starts in Codex or ChatGPT Work draw on your normal allowance, and OpenAI has not said how much proactive research uses.
How far does the free Muse allowance go?
For everyday errands, 100 million tokens a week is enough for most people. Muse counts tokens rather than tasks, so one long browsing job can use more than many short questions. Power at $20 and Maximum at $100 a month raise the weekly cap.