Botpress

What is Botpress?

Botpress is a platform for creating and operating AI agents, particularly agents that talk with customers or employees and use business knowledge and tools. Its product family includes a browser-based visual Studio, a TypeScript agent development kit, integrations, webchat, cloud APIs, and an enterprise support product. A builder can connect documents or structured data, design the conversation and actions, choose supported language models, publish to a website or channel, and inspect activity after launch.

That breadth makes Botpress more than a prompt wrapper, but it does not remove the work of designing a reliable service. A production agent still needs a defined job, approved knowledge, access boundaries, refusal behavior, human escalation, test cases, cost controls, and someone responsible for changes. The official Botpress documentation is the best starting point for checking the current product surfaces and implementation details.

How the building workflow works

In Studio, teams can model an agent with visual flows, reusable instructions, knowledge bases, tables, integrations, and specialist components. Visual editing is useful for collaboration and for making important branches visible. The code-oriented ADK is a better fit when a team wants source control, reusable engineering patterns, or custom behavior that would be awkward to express only through a canvas.

Knowledge bases allow the agent to answer from uploaded or connected material. This is not a substitute for information architecture. Remove outdated duplicates, identify authoritative sources, split content in a retrieval-friendly way, and attach an owner and review date. Test questions should include missing answers, conflicting documents, ambiguous requests, prompt injection, and requests that the agent must escalate.

Integrations let an agent retrieve data or take actions. Give every integration the minimum permissions required. A support bot that only needs an order status should not receive broad write access to the entire commerce account. High-impact actions such as refunds, account changes, messages, or record deletion should have validation, confirmation, audit logs, and often a human approval step.

Botpress also provides webchat and channel deployment, conversation history, analytics, and—in eligible plans—human handoff and expanded collaboration controls. These features help operate an agent, but teams should define service hours and what happens when the human queue is unavailable.

Pricing and the real cost model

The official pricing page separates the platform subscription from AI Spend. The starting pay-as-you-go tier is listed at zero dollars per month and includes a small AI credit, while paid tiers add capacity and operational features. Limits and add-ons may cover incoming messages and events, table rows, bots, collaborators, vector storage, file storage, or reserved responsiveness. Supported model consumption is billed according to the provider price, and Botpress documents workspace spending caps.

Therefore, the headline monthly price is not a complete forecast. Build a representative test set and measure ingestion, model calls, retries, translations, summaries, and average conversation length. Add projected traffic, storage, seats, environments, support, and human-handoff labor. Monitor the dashboard and set conservative spending limits before exposing an endpoint publicly.

Privacy, retention, and security

Botpress's privacy statement distinguishes account and website data from Conversation Data submitted by end users to customer bots. It explains processing, international transfers, service providers, and customer responsibilities. The service terms say Conversation Data and analytics derived from Conversation Data are not used to train or develop algorithms or models, while other analytics may be used to improve the service. Buyers should read the current agreement rather than generalizing from that sentence to every connected model or integration.

The documentation currently describes standard retention examples including 30 days for logs and 90 days for conversations, messages, and events, while files and some other data may remain until deleted. Enterprise options can change retention or residency. Verify the plan-specific configuration, deletion workflow, subprocessors, data-processing agreement, and downstream model policies for the actual deployment.

Security features mentioned by Botpress include encryption, role controls, allowed origins, rate limits, a web application firewall, audit capabilities, and spending limits, with availability varying by plan. The platform cannot secure unsafe workflow code or excessive third-party permissions on behalf of the customer. Treat prompt instructions as untrusted input, validate tool parameters, separate development from production, rotate secrets, and rehearse incident response.

Strengths, trade-offs, and alternatives

Botpress is strongest when a team wants one environment for agent design, knowledge, channels, actions, observation, and escalation, plus a path from visual building to code. That can shorten the distance from prototype to an operated customer experience.

The trade-off is a multi-part system and bill. Model choice, vector storage, messages, add-ons, connected services, and human operations all affect results. No builder automatically guarantees grounded answers or compliant use of personal information.

Compare Dify for an open-source-oriented LLM application platform, Langflow for visual experimentation with model components, n8n for broader workflow automation, or Activepieces for an automation-first approach. Choose after prototyping the same real workflow, including failure and escalation cases—not after comparing feature-count tables alone.