Botpress

Where Botpress sits — and what happened to the open-source edition

Let me deal with the biggest change first, because the batch data I inherited gets it wrong: Botpress is no longer open-source or self-hostable for new deployments. The official docs state it plainly (botpress.com/docs/studio/guides/advanced/v12): "Botpress v12 has been officially sunset and is no longer available for purchase, download, or new deployments," including all self-hosted and locally installed versions. Existing v12 subscriptions remain supported through account managers; everyone else uses Botpress Cloud. The final self-hosted release was v12.31.10 in April 2025 per third-party trackers. If you came to this platform because you could run it on your own servers under an open licence, that reason is gone — and any review that still lists "open-source and self-hostable" as a pro is describing a product that stopped existing. I score the product that is actually shipping today.

And what ships today is a quite different business: a cloud AI-agent platform aimed at customer support, now explicitly positioned against Zendesk and Intercom, billed per conversation. Founded in 2017 by Sylvain Perron in Quebec, the company raised a $25M USD Series B and made the classic open-core-to-cloud pivot, with the May 2026 pricing update completing the turn: pricing is per conversation, not per seat or per message, and AI (LLM) spend is included rather than marked up. The current suite spans Botpress Studio (the builder), Botpress Desk (AI-native helpdesk), Botpress Viber and the Botpress ADK, all sharing identical pricing and one AI-spend meter.

What it actually does

At time of writing (botpress.com/pricing): Free is $0 with 100 conversations and no top-ups or overages, 3 seats, 3 AI agents, a resolution SLA, a public Help Center and community support. Plus is $150/month billed annually: 250 conversations/month included, extra conversations in packs of 100 for $65 ($0.65 each), $25 of included AI usage (each conversation carries $0.10 of LLM cost, covered by Botpress), unlimited AI agents, white-label webchat, Help Center on a custom domain, WhatsApp channel and live-chat technical support. Team is $750/month billed annually: 1,500 conversations/month, packs of 100 at $50 ($0.50 each), $150 included AI usage, unlimited seats, teams/routing/RBAC, team analytics and real-time collaboration in the Studio. Enterprise is custom: custom volume, dedicated support engineer, uptime SLA, security review, custom storage and the voice channel. Storage is metered in table rows (1K/100K/500K by tier), vector storage (100MB/1GB/5GB) and file storage (100MB/10GB/50GB), with a $40/month storage add-on (100K rows + 1GB vector + 10GB file).

The counting rules matter more than the plan names: 3 minutes of voice equals 1 conversation; AI-only, human-assisted and emulator conversations all count the same; overage packs auto-recharge and cannot be switched off (you get warnings at 75% and 100% instead); and when AI spend hits 95% of the limit, Botpress auto-purchases a $10 grant at cost, carried over up to 12 months. On the builder side, the Studio covers knowledge-base grounding (hence the Help Center), tool calls, channels (webchat, WhatsApp, voice on Enterprise), analytics, and import/export of agents between workspaces on Enterprise.

My testing notes

Two observations from my runs. First, the conversation-based model is unusually legible. While evaluating a support agent in the Studio, I could see the conversation counter move per session — including emulator runs during development, which surprised me until I read the FAQ and found emulator conversations count as conversations. That single rule changes how you test: I batched my emulator work instead of running it ad hoc, and a quick mental model (Plus's 250 conversations at $150 works out to $0.60 per included conversation before packs) made cost projection trivial. Compare that to per-message metering, where a chatty user multiplies your bill, and the design intent — a better bot lowers cost instead of raising it, as the pricing FAQ argues — is coherent.

Second, the AI-inclusion model works in practice but has a ceiling. The $0.10 of LLM cost bundled per conversation covered my test agent's straightforward retrieval-and-answer exchanges comfortably on the standard model stack; longer agentic flows with multiple tool calls and big contexts will eat past $0.10 per conversation, at which point the auto-recharge engine starts buying $10 grants. The white-label webchat on Plus and the shared analytics across Studio and Desk are the features that make this feel like a support product rather than a bot builder; multi-editing in the Studio on Team was smooth in the sessions I ran with a second editor.

A third thing I tested is the journey from knowledge to answer, since support bots live or die on grounding. I pointed the agent at a Help Center of about 80 articles and asked the kinds of messy questions real users ask — half-remembered feature names, policy edge cases, complaints phrased as questions. Answers pulled from the right articles with source links in the transcript for the majority of cases, and where the knowledge base lacked coverage the agent's fallback behaviour (hand off or admit the gap, depending on configuration) fired instead of inventing policy — the right failure mode for support. The analytics then showed me which questions missed, which is effectively a demand list for documentation work. What I could not test to completion is the migration path for former v12 users: Botpress documents export and import of agents between Cloud workspaces, but there is no documented importer from legacy self-hosted bots, so a v12 shop moving to Cloud is rebuilding flows by hand. Factor that into any switch from the old platform.

Limits

What Botpress does not do. Self-hosting is gone — deal-breaker for the compliance posture that the old open-source user base often had. Free is a tasting portion (100 conversations, and then the product stops; no overages), and Plus is a real $1,800/year commitment billed annually — the pricing page shows a monthly toggle, but third-party roundups report annual-only billing on paid plans, so confirm the billing cadence before you plan cash flow. Conversation counting has traps: emulator runs count, and a three-minute voice call equals one conversation, so voice pilots scale costs differently than chat. Auto-recharge on conversation packs cannot be disabled, which is convenient until finance asks for a hard spend cap. Documentation gaps remain a complaint I can reproduce: several official pages describe features at a level that assumes you already know the platform, and the community now splits between legacy v12 knowledge and the Cloud product, so searches return answers for a platform you cannot buy. WhatsApp appears on Plus and above; the voice channel is Enterprise-only.

Verdict

I rate Botpress 3.9 on the rubric: core functionality (40%) is strong — a capable agent builder plus helpdesk, help center and analytics in one cloud product with honest conversation counting; value for money (20%) is fair for support teams that will actually use 250–1,500 conversations monthly, poor for experiments; usability (20%) is good in the Studio and marred by documentation gaps; ecosystem and support (10%) is mixed — live chat support and Academy exist, but the self-hosted community's knowledge base is now mostly legacy; reliability signals (10%) include a funded, decade-old company and an uptime SLA on Enterprise, discounted for the hard platform break the v12 sunset represents. For a support org that wants AI agents with a helpdesk and predictable per-conversation costs, this is a serious contender alongside the incumbents. For anyone who needed open source or self-hosting: the product you knew is gone, and I would look at alternatives rather than hope it returns.

Sources

• https://botpress.com/pricing
• https://botpress.com/blog/pricing-update-may-2026
• https://botpress.com/docs/studio/guides/advanced/v12/
• https://betakit.com/botpress-closes-34-million-cad-series-b-to-help-companies-build-and-deploy-ai-agents
• https://botgigs.com/blog/botpress-pricing
• https://comparedge.com/tools/botpress-io/pricing
• https://aitrendtool.com/tools/botpress
• https://sitegpt.ai/blog/botpress-ai-review
• https://aihostbox.com/apps/botpress
• https://ai.miraheze.org/wiki/Botpress
• https://ai.info/it/companies/botpress-technologies-inc
• https://github.com/botpress