What Voiceflow does
Voiceflow is a collaborative platform for building AI customer-experience agents with visual workflows, knowledge, tools, evaluations, and production environments.
Voiceflow gives product, conversation-design, and engineering teams a shared canvas for customer agents. Builders can combine prompts and agentic playbooks with fixed workflows, knowledge retrieval, variables, APIs, functions, and integrations, then test in prototypes before promoting through development, staging, and production environments.
Starter is free and intended for evaluation. Paid plans add models, agents, editor capacity, support, fallback behavior, and enterprise controls; production activity consumes shared credits, while editor seats, phone numbers, concurrency, PII redaction, and extra credit bundles can add cost. The public site now directs buyers to the dashboard or sales for exact current totals, so there is no defensible universal monthly price.
A visual flow does not guarantee a safe conversation. Retrieved documents can be stale or adversarial, model answers can be wrong, and tools can expose data or take unintended action. Voiceflow reports SOC 2 Type II and ISO 27001 controls, encryption, access controls, and enterprise SSO, but deployers still own consent, retention, channel rules, least privilege, evaluation, escalation, and human approval.
How Voiceflow works
Builders provide instructions, conversation paths, approved knowledge, integrations, API or MCP tools, and channel settings. At runtime Voiceflow combines the current conversation with retrieved context and configured model calls, follows deterministic steps where defined, and may invoke permitted tools. It returns a spoken or written response and records traces for review; people must test edge cases and gate consequential actions.
How to set up Voiceflow
Define the service boundary
Choose one intent, audience, channel, handoff rule, approved knowledge set, prohibited data, success metric, and accountable owner.
Build the safest path first
Use deterministic steps for identity, consent, policy, and actions; reserve open generation for bounded explanation and clarification.
Connect only approved context
Upload current sources, attach provenance, scope APIs and MCP tools to least privilege, and keep write actions behind confirmation.
Test conversations and cost
Evaluate normal, ambiguous, multilingual, abusive, privacy, injection, outage, and escalation cases while tracking latency and credits.
Promote with monitoring
Move through environments, review transcripts and evaluations, set usage alerts, audit access, and maintain a staffed human fallback.
Voiceflow FAQs
Is Voiceflow free?
Yes. Starter is free for trying the builder, but its limits and credit allotment are not intended for production.
How does Voiceflow pricing work?
Bills combine a plan, shared usage credits, and optional add-ons such as editor seats, numbers, concurrency, or PII redaction. Enterprise is contracted.
Can Voiceflow build both voice and chat agents?
Yes. Official materials describe deployment across voice and chat, with channel-specific telephony and concurrency considerations.
Can Voiceflow call external systems?
Yes. Agents can use APIs, functions, integrations, and MCP tools when configured. Credentials, permissions, validation, and approvals remain the deployer's responsibility.
Does an agent need human review?
Yes. Teams should inspect traces, evaluate failures, verify knowledge and tool results, and provide escalation for consequential or unresolved cases.
Listing reviewed 2026-07-24. Product details and pricing can change; verify important terms on the provider's website.
Related Automation AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.