Botpress vs Voiceflow

These two are the closest match in this whole set. Both give you a canvas, a knowledge base and a widget at the end of it. The clearest difference between them is not a feature at all. It is that one tells you what it charges and the other does not.

Prices read 18 September 2026

Side by side

Botpress, Voiceflow and Deacon, feature by feature
Feature	Botpress	Voiceflow	Deacon
Free plan	25 conversations, and its own pricing script calls them a one-time allotment rather than a monthly allowance. Three seats, three agents, hard-capped. No card	A plan called Free, described as a one-time credit grant that does not renew and not suitable for production. No figure is published. No card	50 answers a month on one seat, renewing, and no card is asked for
Cheapest paid plan	Plus, $189 a month, or $150 a month billed annually, for 250 conversations	Pro, for individual builders. No price is published on the pricing page	Starter, $50 a month, or $40 a month billed yearly, for 1,000 answers
Where you find the price	On the pricing page, per plan, with the overage rate and the storage add-on beside it	Inside the product. The billing documentation says to open the Plans and Billing tab in your dashboard	On the pricing page
What the meter counts	A conversation. Any exchange with at least two messages from the visitor in the month, with three minutes of a voice call counted as one	Credits, charged per action in dollars. $0.005 an inbound message, the model’s token cost on top, $0.05 a minute for phone	An answer. One question, one reply, one unit
Model spend	Its own allowance. Plus includes $25, every conversation carries $0.10 of it, and top-ups are bought at cost	Metered per thousand tokens, per model, on top of the message charge	Ours. The plan price is the price
When you run out	A pack of 100 conversations is added automatically at 95%. You cannot turn that off, and Botpress answers the question on its own page with the word "No"	Auto top-ups buy more when the balance hits zero, but you have to switch them on. Voiceflow documents phone calls being declined when an organisation is out of credits	The widget carries on, and nothing is bought on your behalf to make it. It asks for an email address so you can follow up
How you build it	Three ways: a drag-and-drop canvas, a conversation with the builder, or a TypeScript library	A canvas for workflows, where every step runs the path you drew, or playbooks that let the model reason toward a goal	You do not. You give it your material
Getting it on your site	Two script tags in the head, once a bot is published	One script block, once an agent is built and an environment published	One script tag before the closing body tag
What it can read	Websites with scheduled re-crawling, documents up to 50 MB, tables you can ask questions of, rich text, Notion and Google Drive, and live web search	Web pages or a whole sitemap, PDFs and Word files up to 10 MB, CSV and Excel tables, plain text, Zendesk, and a Shopify catalogue, refreshed daily, weekly or monthly	Your documentation, pages to crawl, uploaded files, and answers you write by hand
Channels	Webchat, email, WhatsApp, Slack, Discord, Telegram, Instagram, Facebook Messenger and voice	Chat, telephony in and out, SMS through Twilio in beta, and anything on the conversations API. No WhatsApp, Teams, Messenger, Instagram or Telegram	One. The widget on your own site
Seats	Three on Free and three on Plus on the visible card; the page’s own structured data and pricing script both say one on Free. It is headed "priced for the work it does, not the seats it fills"	Editor seats are a separate purchase. Each paid plan includes some, the number is not published, and viewers are free and unlimited	one seat free, three seats on Starter

Free plan

Botpress: 25 conversations, and its own pricing script calls them a one-time allotment rather than a monthly allowance. Three seats, three agents, hard-capped. No card
Voiceflow: A plan called Free, described as a one-time credit grant that does not renew and not suitable for production. No figure is published. No card
Deacon: 50 answers a month on one seat, renewing, and no card is asked for

Cheapest paid plan

Botpress: Plus, $189 a month, or $150 a month billed annually, for 250 conversations
Voiceflow: Pro, for individual builders. No price is published on the pricing page
Deacon: Starter, $50 a month, or $40 a month billed yearly, for 1,000 answers

Where you find the price

Botpress: On the pricing page, per plan, with the overage rate and the storage add-on beside it
Voiceflow: Inside the product. The billing documentation says to open the Plans and Billing tab in your dashboard
Deacon: On the pricing page

What the meter counts

Botpress: A conversation. Any exchange with at least two messages from the visitor in the month, with three minutes of a voice call counted as one
Voiceflow: Credits, charged per action in dollars. $0.005 an inbound message, the model’s token cost on top, $0.05 a minute for phone
Deacon: An answer. One question, one reply, one unit

Model spend

Botpress: Its own allowance. Plus includes $25, every conversation carries $0.10 of it, and top-ups are bought at cost
Voiceflow: Metered per thousand tokens, per model, on top of the message charge
Deacon: Ours. The plan price is the price

When you run out

Botpress: A pack of 100 conversations is added automatically at 95%. You cannot turn that off, and Botpress answers the question on its own page with the word "No"
Voiceflow: Auto top-ups buy more when the balance hits zero, but you have to switch them on. Voiceflow documents phone calls being declined when an organisation is out of credits
Deacon: The widget carries on, and nothing is bought on your behalf to make it. It asks for an email address so you can follow up

How you build it

Botpress: Three ways: a drag-and-drop canvas, a conversation with the builder, or a TypeScript library
Voiceflow: A canvas for workflows, where every step runs the path you drew, or playbooks that let the model reason toward a goal
Deacon: You do not. You give it your material

Getting it on your site

Botpress: Two script tags in the head, once a bot is published
Voiceflow: One script block, once an agent is built and an environment published
Deacon: One script tag before the closing body tag

What it can read

Botpress: Websites with scheduled re-crawling, documents up to 50 MB, tables you can ask questions of, rich text, Notion and Google Drive, and live web search
Voiceflow: Web pages or a whole sitemap, PDFs and Word files up to 10 MB, CSV and Excel tables, plain text, Zendesk, and a Shopify catalogue, refreshed daily, weekly or monthly
Deacon: Your documentation, pages to crawl, uploaded files, and answers you write by hand

Channels

Botpress: Webchat, email, WhatsApp, Slack, Discord, Telegram, Instagram, Facebook Messenger and voice
Voiceflow: Chat, telephony in and out, SMS through Twilio in beta, and anything on the conversations API. No WhatsApp, Teams, Messenger, Instagram or Telegram
Deacon: One. The widget on your own site

Seats

Botpress: Three on Free and three on Plus on the visible card; the page’s own structured data and pricing script both say one on Free. It is headed "priced for the work it does, not the seats it fills"
Voiceflow: Editor seats are a separate purchase. Each paid plan includes some, the number is not published, and viewers are free and unlimited
Deacon: one seat free, three seats on Starter

One of them publishes its prices

Botpress puts almost everything on one page. Plan price monthly and annually, conversations included, the price of a pack when you go over, the model spend included and what a top-up costs, table rows, vector storage, file storage and the storage add-on. It also answers the awkward questions in its own FAQ, including whether you can switch auto-recharge off, where the answer is no.

Voiceflow’s pricing page has no plans and no figures on it at all. It offers to let you start building or to book a demo. The plan names are in the documentation and the billing page tells you to open the Plans and Billing tab inside the product to see what they cost. What Voiceflow does publish is the rate card, in unusual detail: half a cent per inbound message, the model’s token cost per thousand, five cents a minute of phone. So you can price a conversation without being able to price a plan.

There is an older Voiceflow post with a full plan table in it, Pro at $60 a month. It is stamped April 2025 and it describes the free tier under a name and a shape the current documentation contradicts, so it is about seventeen months out of date. A couple of 2025 figures also survive in the documentation changelog. None of that is current pricing and none of it should be treated as a quote.

Which of the two to pick

Voiceflow is the stronger voice product. Telephony runs in both directions, speech to text and text to speech are each metered and each swappable, and the workflow model is built for an agent that has to take the same path every time. If what you are making is a phone agent, this is the one.

Botpress is the stronger chat product and the easier one to budget for. Nine channels against four, including WhatsApp, Telegram, Instagram and Messenger, which Voiceflow does not carry at all. The conversation meter does not move with model choice, the model spend is sold at cost, and three seats come with every plan including the free one. You can also skip the canvas entirely and write the agent in TypeScript.

Both of these are workbenches, and a workbench is only worth the bench fee if you were going to build something. If every answer your visitors need is already written down on your site, there is nothing to draw and nothing to publish. Deacon takes the pages you already have and answers from them.

Where these figures came from

Every figure above was read on 18 September 2026, from these pages.

Try Deacon for free

If the answer to every question is already written down somewhere on your site, there is nothing to build. Deacon reads it and answers from it.

The Deacon and ChatBot comparison is the one about drawing branches by hand. If you already know you want the no-canvas route, how to use Deacon walks the install, and the pricing page has the figures.