Botpress alternatives in 2026: Voiceflow, Chatbase and 4 others compared

The ranking, from closest to what Botpress does to most specialised:

1. Voiceflow to design voice and chat agents with fine control.
2. Chatbase for an agent live in a day from your website.
3. Copilot Studio if you already run Microsoft 365 and Azure.
4. n8n for an agent built into your workflows, at the lowest cost.
5. Fin for high-volume support, billed per outcome.
6. Crisp for customer messaging from a French vendor, with an AI agent, at a flat price.

Outside the ranking, if you want the agent built and maintained for you: our custom AI agent offer.

This guide only compares Botpress replacements. For ChatGPT Enterprise, Claude, Mistral, Zendesk or Tidio, see our comparison of AI chatbots for business. Disclosure: Lumyniq publishes this site and sells the custom agent offer shown outside the ranking.

Three things changed, according to the 14 May 2026 post. Botpress used to bill each incoming message; it now bills the whole conversation. It used to bill AI usage separately, at the provider’s price with no markup; AI is now included in each conversation you buy. And bots, which were a $10-a-month add-on each, are unlimited on paid plans.

The overage mechanics are worth reading in full. The pricing FAQ explains that at 95% of quota a pack of 100 conversations is added automatically, that you are notified at 75% and 100%, and that packs persist from month to month until you remove them. If AI usage reaches 95% of its allowance, Botpress also buys $10 of AI credit, sold at cost. The Free plan stops at 100 conversations.

Botpress argues against the outcome billing used by Fin and Zendesk: in its words, outcome pricing raises the bill as the bot gets better. In return, under per-conversation billing you also pay for the exchanges the bot failed to handle.

What it does well. The bill is predictable: you know the price of a conversation before the month starts, with no token maths. Agents are unlimited from Plus, so you can run one agent per job (support, booking, qualification). Storage was raised in May 2026: 100,000 table rows, 1 GB of vectors and 10 GB of files on Plus, 500,000 rows, 5 GB and 50 GB on Team. White-label webchat and WhatsApp are among the listed channels.

Where it falls short.

• Entry price. The first paid tier costs $189 a month for 250 conversations, or $0.76 per conversation before the first pack. Chatbase Hobby costs $40.
• Test conversations count. The pricing FAQ says emulator conversations are counted like any other. A team that tests a lot burns quota before go-live.
• Human conversations count too. An exchange taken over by an agent in Botpress Desk is billed like one handled by the bot.
• The 95% trigger. Because a pack is added before the quota is reached, the bill can include one pack more than strictly needed. In our simulation, 9 packs instead of 8 at 1,000 conversations on Plus.
• Hosting. According to its privacy statement updated on 17 August 2026, AWS hosts Botpress’s cloud services and all customer data, and conversations are stored on servers in the United States or other locations set by the customer. Transfers from the EU rely on the Data Privacy Framework. We found no published European region. Ask for the data processing agreement in writing.
• Self-hosting. Botpress Cloud does not run on your servers. The self-hostable version, Botpress v12 under AGPL-3.0, lives in a separate repository; the main repository now holds only the integrations and developer tools.

Botpress also sells Built For You, a service where its team builds and maintains the agent, listed at $1,650 a month in its May post. It answers the same need as our custom AI agent offer, with a different model: a monthly subscription tied to the Botpress platform, against setup billed once, a maintenance plan, then the model paid at API cost.

This counts only the packs strictly needed. With the 95% trigger, the twelfth pack arrives at 1,283 conversations, so the crossover falls to about 1,280. Above 1,500 conversations, the gap widens in Team’s favour by $15 per pack. If your volume regularly exceeds 1,350 conversations (about 1,280 with the 95% trigger), Team costs less and its seats are unlimited.

The most useful gap to read is the billing unit: Botpress and Crisp count conversations, Chatbase and Copilot Studio count replies, n8n counts executions, Fin counts outcomes. An agent that asks many follow-up questions costs more on Chatbase or Copilot Studio and the same on Botpress. An agent that resolves well costs more on Fin and the same elsewhere.

The table also compares tools that do different jobs. Crisp and Fin include an inbox for your team, n8n includes neither a support interface nor build time, and Copilot Studio assumes the Microsoft environment. Add build and maintenance time to every line.

• Botpress. 1,000 conversations on Plus: 250 included, 8 packs of 100 at $65, so $189 + $520 = $709. 5,000 on Team: 1,500 included, 35 packs at $50, so $939 + $1,750 = $2,689. On annual billing: $670 and $2,500. With the 95% trigger, count $774 and $2,839.
• Chatbase. 4 credits per conversation on a 1-credit model. 4,000 credits fit in Basic ($150). 20,000 credits: Business (15,000 credits, $500) plus 5,000 credits at $40 per 1,000, so $700.
• Copilot Studio. 4 generative answers at 2 credits, so 8 credits per conversation. 8,000 credits at $0.01 pay-as-you-go: $80. 40,000 credits: one 25,000-credit pack at $200 plus 15,000 pay-as-you-go, so $350, provided pay-as-you-go is enabled on the environment. Without it, Microsoft Learn says custom agents are disabled above 125% of prepaid capacity. Each agent action adds 5 credits.
• n8n Cloud. Each message runs an execution, so 4,000 then 20,000 executions: Pro 10,000 at €60, then Pro 50,000 at €144. The Pro selector offers no tier between 10,000 and 50,000. Mistral Large model at the price published on mistral.ai/pricing ($0.5 per million input tokens, $1.5 per million output): $0.0078 per conversation, so €7 then €35.
• Voiceflow. 4 messages at $0.005 plus gemini-2.5-flash tokens at the credits-table price: $0.0282 per conversation, so $28 then $141. The plan fee is not published, nor is the share of usage covered by included credits.
• Fin. $0.99 per outcome. We assume 50% of conversations resolved or handed off as configured, and give the ceiling at 100%. Your helpdesk cost is not counted.
• Crisp. Essentials at €95 with €25 of AI credits, plus Hugo at the top of Crisp’s own range, $0.10 per conversation: $100 then $500 of AI, minus the included credits. The pricing page, read from France, shows prices and credits in euros; Crisp’s help center states credits in dollars. We use Essentials because Mini (€45) includes the AI Agent but neither AI training on your internal data nor the knowledge base, according to the pricing page comparison. With Mini, the calculation gives €130 then €489.

The ECB had not yet published the 29 October 2026 rate at the time of calculation (it publishes around 16:00 CET), so we used the 28 October rate. Change one assumption and the order moves: with 8 replies per conversation, Chatbase and Copilot Studio roughly double ($310 and $160 at 1,000 conversations), Botpress stays the same.

A French option outside the ranking: Dydu. Dydu has built chatbots, voicebots and callbots since 2009 and states that its data is hosted in France. Its prices are not published (demo on request), so we could not price it. Worth a look if hosting in France is a contractual requirement.

For n8n, the licence question and the alternatives to n8n itself are covered in our guide to n8n alternatives by licence. If the agent mainly needs to trigger automations between your tools, our comparison of Zapier alternatives covers the other half of the subject. To build an agent yourself on one of these tools, see our guide to building an AI agent without code.

To place the agent in a wider project (definition, use cases, cost of a conversation), see our guide to AI agents for business.

If what you want is a self-serve tool where you paste your website address and get an agent: Lumyniq is preparing Agent Studio, waitlist open. It is not available yet, so it is not part of this comparison (the name is close to a Botpress module, with no connection to it).

Quality. A cheaper agent that hands twice as many conversations to a human can cost more in total. Test two tools on your fifty most frequent questions before choosing.

Assumptions. Four replies per conversation and 3,000 tokens read per reply are orders of magnitude. An agent that reads many documents or chains actions consumes far more on tools billed per reply or per token.

Data. Price says nothing about hosting. Crisp states that it stores data in the EU. Botpress stores conversations in the United States or other locations set by the customer, according to its privacy statement of 17 August 2026. We could not verify Voiceflow’s region outside its Enterprise plan. For personal data, ask for the data processing agreement and the list of subprocessors.

Migration. None of these tools imports a Botpress bot. Plan for re-importing the knowledge base and rebuilding flows and integrations, to be costed by the number of flows and integrations.