Free for one site. An espresso a month for everything else.

The other cost: your AI provider

AgentPresso is the brain — all the WordPress skills, the safety model, the tool execution, every workflow already loaded. The AI model is the fuel: it understands what you say, you pick the supplier.

Token pricing is split into input (what the model reads — chat history, tool output, your message) and output (what the model writes back). With an agent, output carries more weight than in plain chat: the model writes its reasoning, decides which tool to call, then writes the result — round after round. And since output tokens cost several times more than input tokens at every provider, the output rate is the number that actually drives your bill.

Ordered from free to most premium:

ChatGPT / Codex — use the subscription you already have

If you already pay for ChatGPT, you can use that existing subscription with AgentPresso. Connect ChatGPT / Codex and choose gpt-5.5 if you want one of the most powerful AI models on the internet. AgentPresso can use the same AI plan you already rely on — no separate API billing just to get started.

LM Studio — completely free

Runs an open-source AI model locally on your Mac. Download once (~4 GB for a good model), point AgentPresso at it. No API key, no tokens to count, no cap, ever. The trade-off is smaller models than the cloud ones — slightly slower or less precise on hard tasks. For routine WordPress maintenance, plenty.

DeepSeek — cheapest serious cloud option

Their deepseek-v4-flash costs roughly $0.30 per million input tokens and $1.10 per million output tokens. A serious agent conversation — output-heavy — lands at roughly 1–2 cents per chat. A $5 starting credit lasts months for most people. Fast, accurate, kind to the wallet.

For heavier work, deepseek-v4-pro runs at roughly $0.80 per million input tokens and $3 per million output tokens — about 4–5 cents for the same kind of conversation. Still inexpensive next to OpenAI or Anthropic, with noticeably stronger reasoning on complex jobs. In our own team we use Flash for almost everything, and only reach for Pro on the genuinely hard tasks.

Google AI Studio (Gemini) — generous free tier, no card

Gemini 2.0 Flash has a free quota that covers most casual use — just sign up with a Google account. Good for trying AgentPresso without spending anything on AI tokens.

If you want Google's premium model, gemini-3.1-pro costs roughly $1.25 per million input tokens and $10 per million output tokens — output-heavy agent chats land around 10–15 cents each. Comparable to Claude on price, with strong long-context performance. Another solid pick when the work is serious.

OpenRouter — one key, many models, including several that are completely free

Useful for experimenting across providers without managing multiple accounts. Free open-source models (Llama, Mistral, Gemma, others) are listed at the filtered URL above. Paid models on OpenRouter cost roughly what they cost on the providers' own platforms, plus a small surcharge.

OpenAI — the most familiar name

GPT-4o costs roughly $2.50 per million input tokens and $10 per million output tokens. Output-heavy agent chats land around 10–20 cents each, depending on how much the model writes back. Still cheap in absolute terms, and the dashboard is the most beginner-friendly of any provider. $5 starting credit lasts weeks of regular use.

Anthropic (Claude) — the premium choice for serious work

Claude Sonnet 4 costs roughly $3 per million input tokens and $15 per million output tokens. That's the most expensive option on this list — agent chats typically run 20–30 cents each. So why pick it? Claude is the model that thinks longest before doing anything destructive, asks the smartest follow-up questions, and writes the cleanest content. When the task matters — a security audit on a compromised site, a complex WooCommerce migration, the content for a flagship launch — Claude pays for itself by being right the first time. Not for every chat, but the model you reach for when the stakes are real.

The point: pick whichever feels right for your kind of work. Use DeepSeek for everyday maintenance, switch to Claude for hard tasks, run LM Studio when nothing should leave your Mac. AgentPresso lets you change provider in two clicks — the freedom to pick is the actual feature.

See Getting started for setup walkthroughs of every supported provider.

Pricing FAQ

How does the Multisite subscription work?

Multisite is billed monthly through Freemius. You enter your card, get a license key in return, and paste it into Settings → License inside AgentPresso. The key works on every Mac you own. You can cancel any time from your Freemius account — when the subscription lapses, AgentPresso returns to the free tier (one site).

Is there a yearly discount?

Yes, but it varies. Check the Freemius checkout for the current annual option — it's usually 15–20% cheaper than paying monthly.

Can I get a refund?

Yes. If AgentPresso isn't what you hoped in the first 7 days, send us a note through the contact form and we'll refund 100% of your money — no questions asked.

What happens to my sites if my subscription lapses?

Nothing dangerous. The app keeps every site connection you've added, but only the first site on your list stays active — the rest are disabled until you renew. No data is deleted, no credentials are lost; renew the subscription and everything switches back on exactly where it was.

Do I need to keep paying for the AI provider too?

Yes, separately, if you use a cloud provider. Or no, if you run a local model. AgentPresso doesn't charge for AI usage — your provider does. We don't see what you pay.

Team licenses?

Volume discounts are available for teams and agencies — three standard tiers at 10, 100, and 1,000 licenses, with the per-seat price dropping at each step.