Botpress vs Voiceflow in 2026: Pricing, AI Agents and Which to Pick

Botpress and Voiceflow are the two names that come up most often when a team decides to build its own AI agent rather than buy a finished one. Both give you a visual builder, a knowledge base, a choice of large language models and ways to put the agent in front of customers. They differ in who they are built for, which channels they reach and, most of all, how they charge.

An AI agent builder is a platform where you design the agent yourself: its logic, the knowledge it searches, the tools it calls and the channels it runs on. That flexibility is the point, and it is also the cost, because someone has to build, test and maintain the agent after launch.

This comparison is for the developer, CX lead or agency who will own that work, and for the founder deciding whether to take it on. It ends with when you don't need a builder at all. Every price below was read from Botpress's and Voiceflow's own pages and documentation on October 10, 2026.

Botpress vs Voiceflow at a glance

Botpress is a developer-friendly agent platform with published, conversation-based pricing and a long list of messaging channels. Voiceflow is a CX-team platform built around chat and voice, with credit-based usage and plan prices you request from sales.

Sources: the Botpress pricing page, the Voiceflow billing documentation and Voiceflow's credits pricing table, read on October 10, 2026.

What is the difference between Botpress and Voiceflow?

Botpress leans toward developers and messaging apps; Voiceflow leans toward CX teams, the phone and governance. The deciding factors are your channels, your team and whether you need a price before a sales call.

Botpress lets you build on a visual canvas in its Studio or entirely in code with its Agent Development Kit (ADK), and its Hub lists fourteen channel integrations, from WhatsApp and Telegram to Slack and WeChat. Botpress Desk, an AI helpdesk, comes on the same plans, and every self-serve price is published.

Voiceflow puts the phone on equal footing with chat, with inbound and outbound calling. It treats the agent as software with a release process, sells to agencies with multi-client workspaces and white-labeling, and lists SOC 2 Type II, ISO 27001 and HIPAA on its security page. What it does not publish is a plan price.

How much do Botpress and Voiceflow cost?

Botpress publishes its plan prices and charges per conversation, with model usage bundled in; Voiceflow publishes its usage rates but not its plan prices, so you can estimate the variable part of the bill and must ask sales for the fixed part.

Botpress: conversations, with AI usage bundled

Botpress's unit is the conversation, which its pricing FAQ defines as "Any exchange with at least two messages from the end user in the billing month." Two rules make the count larger than you might expect: "A conversation that spans two months counts in both months," and "AI-only and human-assisted conversations are counted the same, as well as emulator conversations." Chats your people handle, and your own test chats, use the same allowance.

As of October 2026 the plans are:

• Free: $0 for 100 conversations, 3 seats and 3 AI agents, with "No top-ups or overages." Unlike the paid cards, the Free card does not say the 100 renews monthly.
• Plus: $150 a month billed annually for 250 conversations a month, $25 of AI usage, 3 seats, unlimited AI agents, WhatsApp, white-label webchat and a Help Center on your own domain.
• Team: $750 a month billed annually for 1,500 conversations a month, $150 of AI usage, unlimited seats, teams and role-based access, routing and assignment rules, and team analytics.
• Enterprise: custom pricing, adding dedicated support, an uptime SLA, a security review, custom storage and a voice channel.

Annual billing saves 20%, and the page text shows only annual prices. Past the allowance, "Once you reach 95% of your quota, a pack of 100 conversations is automatically added to your account," at $65 on Plus and $50 on Team. You cannot switch that off; the FAQ explains that "Pausing conversations means pausing the product entirely." Packs persist month to month until you remove them, and unused conversations expire at month end.

Model costs are mostly hidden inside the conversation price: "Every conversation includes $0.10 of AI usage, covered by us." If your model calls cost more, a $10 credit grant is bought at cost each time AI spend reaches 95% of the limit. The same prices apply to Studio, Botpress Desk and the ADK.

Voiceflow: credits, with plan prices behind a form

Voiceflow's pricing page lists two options and no prices. "For Agencies & Partners" offers a "Free trial, no credit card required" and "Transparent, usage-based billing"; "For Businesses" offers "Book a demo" and "Request pricing." The billing documentation describes four plans: Free, with "a one-time credit grant that does not renew, so it isn't suitable for production usage"; Pro for individual builders; Business for growing teams, with fallback models and priority support; and Enterprise, with single sign-on, private cloud hosting and custom credit allotments. The plan comparison lives inside the product.

A credit is Voiceflow's unit of usage, and the credits table prices it in dollars. As of October 2026, every user message costs $0.005, phone calls cost $0.05 a minute, and model calls are priced per 1,000 tokens: for example $0.000188 in and $0.00075 out for gpt-4o-mini, and $0.00375 in and $0.01875 out for claude-sonnet-4-6. Paid plans get a monthly allotment that resets each cycle, and auto top-ups "automatically purchase additional credits when you hit zero."

Seats are simpler: every paid plan includes some editor seats, more can be bought, and "Viewers are free and unlimited on all plans." One Voiceflow phone number per workspace is included. The last public plan prices came from Voiceflow's April 2025 credits announcement, Pro from $60 a month and Business from $150, with extra editor seats at $50; treat those as history and confirm the current figures in the product.

Worked example: 1,000 conversations a month

At 1,000 support conversations a month, Botpress costs roughly $670 to $750 a month on annual billing, while Voiceflow's usage comes to about $23 to $76 in credits on top of a plan price it does not publish.

The assumptions, so you can swap in your own: 1,000 conversations a month, each with four visitor messages and four AI replies, and each reply reading about 3,000 tokens of instructions and retrieved text and writing about 150.

Botpress needs eight or nine packs because a pack is added whenever you reach 95% of the current quota, so at this volume Team, with unlimited seats and routing, costs about the same as Plus. On Voiceflow the subscription is the unknown; get the plan price, included credits and seat price in writing before you build. On both, the largest cost is on neither invoice: the hours spent building and maintaining the agent. The AI chatbot cost guide compares these pricing models with per-resolution and flat-plan vendors.

Which is easier to build with, Botpress or Voiceflow?

Neither is a weekend job. Voiceflow steers you toward an agentic framework of playbooks and workflows; Botpress gives you a canvas of nodes with an LLM-driven Autonomous Node and a full code path through its ADK. Which is easier depends on whether your builder thinks in prompts and diagrams or in code.

In Botpress Studio, a standard node executes its cards one by one, while the Autonomous Node uses a large language model "to decide when to execute tools," configured with instructions, variables, tools, knowledge base search, workflows and exit conditions. Actions are reusable code snippets the Autonomous Node can call, and the docs recommend a high-performing model, roughly GPT-4.1 class, to power it. Developers who prefer files to canvases can use the ADK, a command-line kit where adk init creates a project, adk dev runs it with hot reloading and adk deploy uploads it to Botpress.

Voiceflow asks you to choose a framework when you create a project. In the agentic framework, "The agent uses playbooks, workflows, tools, and reasoning to navigate conversations autonomously"; in the conversation flow framework, "You design the conversation path in the visual builder, with or without AI." The docs advise, "If you're unsure, start with the Agentic framework." Around the canvas, its documentation covers environments, A/B testing, merging and version history, plus evaluations, tests and analytics once the agent is live.

Teams used to staging and rollbacks will find Voiceflow's release tooling familiar; teams that prefer code to diagrams will find the ADK more natural.

Which channels can Botpress and Voiceflow deploy to?

Botpress covers messaging apps; Voiceflow covers the phone. That single difference settles the choice for many teams.

The Botpress Hub lists channel integrations for Webchat, WhatsApp, Messenger and Facebook, Instagram, Telegram, Slack, Microsoft Teams, Discord, LINE, Viber, WeChat, Twilio, Vonage and Sunshine Conversations. WhatsApp and white-label webchat are Plus features on the pricing page, and a voice channel is listed on Enterprise.

Voiceflow's deployment documentation covers a chat widget, phone, SMS and an API. It "can provision a US or Canadian phone number for you," or you can connect your own number from Twilio, Vonage or Telnyx, with inbound and outbound calling. SMS is marked beta. The deploy section does not list WhatsApp or Messenger, so a messaging-app channel means connecting Voiceflow's API to a third-party provider.

If customers phone you, Voiceflow is the stronger tool; if they write on WhatsApp, Telegram or Slack, Botpress is. For what AI can and cannot do on a phone line today, see our guide to AI voice agents for customer support.

How do their knowledge bases compare?

Both answer from content you supply. Botpress accepts larger files and more source types, including live web search; Voiceflow imports from Zendesk and Shopify and lets you choose how often each source refreshes.

According to Botpress's documentation on adding knowledge base sources, a knowledge base can hold websites, either a root domain whose subpaths you pick from or specific pages; documents in PDF, HTML, TXT, DOC, DOCX and Markdown "up to 50 MB each"; tables; rich text written in Botpress; web search, which queries the open web live; and content from integrations such as Notion or Google Drive. You can set a recrawl schedule for selected pages. Storage is capped per plan, with a $40-a-month add-on on Plus and Team.

Voiceflow's knowledge base takes publicly accessible web pages, sitemaps, PDF, TXT and DOCX documents up to 10 MB, CSV and XLSX tables up to 10 MB with each row becoming a chunk, plain text, Zendesk help-center articles and Shopify product catalogues. URL and integration sources refresh never, daily, weekly or monthly, and a lower refresh rate saves credits when chunking uses a model.

On either platform, test what the agent says when the answer is not in your knowledge base before you launch.

Human handoff, team features and security

Botpress hands conversations to its own helpdesk; Voiceflow hands them to the contact-centre platform you already run. Voiceflow publishes more security certifications; Botpress publishes more of its team features and prices.

The Botpress human-in-the-loop documentation requires "A Botpress Plus plan" or higher and notes that "For new workspaces, HITL has moved to Botpress Desk, our new AI helpdesk." Desk lists priorities, custom statuses, tags, macros and AI routing with capacity-based assignment and business hours. The Hub also lists handoff integrations for Intercom, LiveChat, HubSpot, Freshchat, Chatwoot, Zoho SalesIQ and Brevo.

Voiceflow's handoff step passes conversations "through native integrations to contact center systems": Genesys, Ujet, Kustomer, Dixa and Zendesk, with others added on request. It relays messages and files both ways and can return the customer to the AI afterwards. Zendesk handoffs need Zendesk messaging enabled, because conversations arrive in its Agent Workspace. There is no Voiceflow inbox of its own, so a small team without a contact centre has to buy one to receive handoffs.

On security, Voiceflow's security page lists SOC 2 Type II, ISO 27001, HIPAA and GDPR, single sign-on and encryption at rest with customer-managed keys, and Enterprise adds private cloud hosting. Botpress's pricing page lists an uptime SLA, a dedicated security review and custom storage on Enterprise; check its trust documentation for certifications before procurement asks.

Who should choose Botpress and who should choose Voiceflow?

Choose Botpress for messaging channels, published prices and code; choose Voiceflow for the phone, release governance and certifications.

Choose Botpress when

• Customers reach you on WhatsApp, Telegram, Slack, Microsoft Teams or other messaging apps.
• You want to know the price before you talk to sales, and a per-conversation bill suits your budgeting.
• Your developers want to build the agent as a code project with the ADK.
• You want the AI agent and the helpdesk in one product with Botpress Desk.

Choose Voiceflow when

• The phone is a primary channel and you need provisioned numbers or Twilio, Vonage or Telnyx.
• You want environments, A/B tests, version history and evaluations around every release.
• Procurement asks for SOC 2 Type II, ISO 27001 or HIPAA.
• You already run Genesys, Kustomer, Dixa or Zendesk messaging and want handoffs to land there.

Do you need an AI agent builder at all?

Not if the job is a support agent on your website that answers from your own content and hands off to your team. A builder earns its cost when the logic, the channels or the governance are genuinely custom.

A finished support agent is a product where retrieval, handoff, the team inbox and the common integrations already work; you supply content and instructions instead of building flows. Our guide to what an AI support agent is explains the category, and the build vs buy guide works through what a built agent costs to maintain.

In a builder, a website support agent needs a synced knowledge base, a prompt that refuses to guess, a handoff path, an inbox, order or ticket lookups and a way to test changes. Each piece is buildable, and each is a piece someone maintains.

PepoChat, which we make, is one finished option. It crawls up to 200 pages of your site, imports sitemaps, documents and Notion or Google Drive pages, answers only from that knowledge base, and hands the conversation to a built-in team inbox, emailing your team, when it finds nothing relevant or the visitor asks for a person. Prebuilt actions check Shopify and WooCommerce orders, look up Stripe subscriptions and file helpdesk tickets. Every feature is on the free plan, with 200 AI replies and 5 knowledge sources a month. Starter is $49 a month for 2,000 replies and 3 members, and Growth is $149 for 10,000 replies and 10 members, with replies past the allowance at $10 per 1,000; the pricing page has the full table. On the worked example above, 4,000 AI replies cost $69 a month on Starter, and your team's replies are free.

The limits are where a builder wins. PepoChat is a website widget and hosted chat page only, with no WhatsApp, Telegram, Slack, phone or SMS channel; its optional voice runs on your own Vapi account, apart from the chat agent's knowledge base and inbox. It runs on OpenAI models with no model menu, has no flow canvas, environments, A/B tests or version history, and publishes no SOC 2 or ISO certification. If your agent must approve refunds in three branches, answer calls or run on WhatsApp, use Botpress or Voiceflow.

What to do next

Write down three things before you sign up anywhere: the channels your customers use, the logic the agent must run beyond answering questions, and the person who will maintain it. If you need a builder, build the same small agent on Botpress's free tier and Voiceflow's trial, with three real questions and one handoff, and compare the hours as well as the price. The Botpress alternative and Voiceflow alternative pages compare each platform with a finished agent feature by feature.

If a finished agent fits, start a free PepoChat workspace, add your site as one knowledge source and run it beside your current setup for a month before paying anyone.