Structured customer support
Route callers through identification, intent capture, troubleshooting, account actions, escalation, and wrap-up.
Independent tool overview
Agent Workflows at a glance
Agent Workflows is ElevenLabs' visual graph editor for routing voice and text conversations inside ElevenAgents. Teams can combine specialized subagents, required tool calls, conditional branches, transfers, and end states, then manage the workflow through the dashboard, CLI, SDK, or API.
Visit the official Agent Workflows site ↗
- Product
- A workflow builder inside ElevenAgents
- Primary interface
- Visual graph editor in the ElevenLabs dashboard
- Configuration path
- conversation_config.workflow
- Management options
- Dashboard, CLI, SDK, and API
- Starting price
- Free with 15 included call minutes
- Key cost boundary
- LLM and telephony charges are separate
Overview
What Agent Workflows is
Agent Workflows is a feature of ElevenAgents rather than a separate automation product. It replaces one large linear prompt with an explicit conversation graph: each phase can use different instructions, models, voices, tools, and knowledge, while edges determine where the conversation goes next.
The visual structure makes complex call flows easier to inspect, but it does not make them deterministic. Natural-language edge conditions are still evaluated by an LLM, tools and integrations can fail, and spoken input can be transcribed incorrectly. Production teams need versioning, repeated tests, safe fallback paths, cost controls, privacy settings, disclosure, and human escalation.
Use cases
Who Agent Workflows is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Lead qualification
Use dedicated phases for consent, qualification questions, CRM updates, scheduling, and human handoff.
Appointment workflows
Coordinate information gathering, calendar lookup, confirmation, rescheduling, and transfer paths.
Multi-specialist voice agents
Assign different prompts, models, voices, tools, and knowledge to distinct conversation phases.
Teams that need observable branching
The graph and per-node analytics are useful when operators need to see where conversations enter, stall, terminate, or change paths.
Capabilities
Core Agent Workflows features
Visual workflow graph
Build and connect conversation nodes in the dashboard instead of encoding every path in one prompt.
Subagent nodes
A node can append to or replace the base configuration, including system prompt, model, voice, tools, and knowledge sources.
Dispatch tool nodes
Create a dedicated point that calls a specific tool, with separate success and failure routes based on its result.
Agent transfer nodes
Hand a conversation from one configured AI agent to another.
Human phone transfer
Route a voice call to a telephone number when a person should take over.
End nodes
Terminate a call at an explicit point in the graph.
Conditional edges
Forward and backward edges can use natural-language LLM conditions, expressions, or unconditional routing.
Code-managed configuration
Pull, edit, and push workflow JSON through the CLI or update it with the SDK and API.
Workflow analytics
The analytics graph overlays node entries, average time, terminations, and incoming and outgoing edge distribution.
Automated agent tests
Simulation, next-reply, and tool-call tests cover full conversations, response criteria, and expected tool parameters.
Probabilistic test runs
Repeat the same test multiple times to measure pass rate instead of relying on one favorable generation.
Versioned rollout
Immutable versions, branches, traffic splitting, merging, and rebasing support safer staged changes.
Process
How the Agent Workflows workflow works
Step 1
Map outcomes and escalation
Define user intents, required disclosures, allowed actions, success criteria, failure states, human handoffs, and emergency routes.
Step 2
Build the smallest useful graph
Create only the nodes needed for the core path, with explicit success, error, timeout, refusal, and transfer behavior.
Step 3
Restrict each node
Give subagents the minimum prompts, knowledge, tools, credentials, and data needed for that phase.
Step 4
Test normal and adversarial cases
Use simulation, next-reply, and exact tool-parameter tests across accents, silence, interruptions, prompt injection, ambiguity, and integration failures.
Step 5
Version and stage traffic
Create an isolated branch, review the full configuration diff, then send a small percentage of traffic to the new version.
Step 6
Monitor and improve
Track node entries, exits, transfers, costs, latency, tool errors, complaints, privacy incidents, and real conversation failures.
Cost
Agent Workflows pricing and free plan
The Workflow Builder is included with ElevenAgents plans. Plans include call minutes and concurrency, while additional hosted call minutes are $0.08 each, burst minutes are $0.16, and text messages are $0.003. LLM usage and telephony-provider charges are billed separately, so the subscription price is not the total operating cost.
Free
$0/month
Evaluation plan with the Workflow Builder.
- 15 included call minutes
- 4 concurrent calls, knowledge base, multilingual support, and widget
Starter
$6/month
Entry commercial plan.
- 75 included call minutes and 6 concurrent calls
- Adds text messages and a commercial license
Creator
$22/month; first month advertised at $11
Higher-usage individual plan.
- 275 included call minutes and 10 concurrent calls
- Supports additional paid minutes
Pro
$99/month
Professional plan with more included usage.
- 1,238 included call minutes
- 20 concurrent calls
Scale and Business
$299 or $990/month
Business plans with more minutes, concurrency, and seats.
- Scale includes 3,738 minutes, 30 concurrent calls, and 3 workspace seats
- Business includes 12,375 minutes, 40 concurrent calls, and 10 workspace seats
Enterprise
Custom pricing
Contract plan for higher scale and governance.
- Elevated concurrency, custom SSO, DPA and SLA terms, and priority support
- BAAs are available for eligible HIPAA customers
Pricing checked . Check current pricing at the source ↗
Assessment
Agent Workflows strengths and limitations
Where it stands out
- Makes branching voice-agent behavior visible as a graph
- Supports specialized prompts, models, voices, tools, and knowledge by phase
- Dedicated tool nodes create explicit success and failure paths
- Built-in AI-to-AI and AI-to-human transfer nodes
- Workflow JSON can be managed with code and CI/CD
- Per-node analytics help locate drop-off and routing problems
- Three complementary automated test types
- Multi-run testing measures probabilistic reliability
- Version branches and traffic percentages support gradual releases
- The Workflow Builder is included on the Free plan
What to consider
- Agent Workflows is not a general business-process automation platform; it is tied to ElevenAgents conversations
- LLM-based edge conditions remain probabilistic and can route the same intent differently
- A dispatch node guarantees that a tool is called, not that the external action succeeds or is correct
- Tool failures, slow APIs, malformed responses, expired credentials, and partial writes need explicit recovery design
- Speech recognition can mishear names, account numbers, dates, amounts, addresses, and consent
- Subagent overrides can accidentally expose broader tools or knowledge than a conversation phase requires
- A visually tidy graph can still hide unsafe prompts, permissions, credentials, or data flows
- Automated tests use models and cannot replace human review, production monitoring, or legal validation
- System and workflow tools are not mocked by the current simulation-test mocking feature
- A single passing test does not establish reliability; repeated tests and real failure cases are necessary
- The plan fee excludes separate LLM and telephony-provider usage
- Additional and burst minutes can make high-concurrency traffic more expensive than the base plan suggests
- Versioning is opt-in and cannot be disabled after it is enabled
- Traffic splitting reduces rollout risk but does not automatically roll back unsafe behavior
- ElevenLabs requires notice that users are interacting with AI and that conversations are recorded and may be shared with service providers
- Recording, consent, telemarketing, accessibility, privacy, retention, and industry rules vary by jurisdiction and use case
- Enterprise-only controls include conversation-history redaction, and advanced contractual protections require the appropriate plan
- Human escalation must remain available for emergencies, disputes, vulnerable users, and consequential decisions
- No agent should autonomously make irreversible financial, medical, legal, employment, safety, or account-security decisions without deterministic controls and authorization
Compare
Agent Workflows alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Sales
Vapi
Consider Vapi for a developer-oriented voice-agent orchestration layer with a different provider and component model.
Explore Vapi →Business Operations
Retell AI
Consider Retell AI for another managed platform focused on low-latency conversational phone agents.
Explore Retell AI →Sales
Bland AI
Consider Bland AI for phone automation with its own pathway and transfer tooling.
Explore Bland AI →Business Operations
Synthflow AI
Consider Synthflow when a no-code sales and lead-conversion phone-agent workflow is the priority.
Explore Synthflow AI →Questions
Agent Workflows FAQs
What are ElevenLabs Agent Workflows?
They are visual conversation graphs inside ElevenAgents. Nodes can represent specialized subagents, required tool calls, transfers, or call endings, while edges control routing.
Is Agent Workflows a standalone product?
No. It is part of the ElevenAgents platform and is included as the Workflow Builder in current agent plans.
Can workflows be version-controlled?
Yes. Workflow configuration lives in the agent configuration and can be managed through the CLI, SDK, and API. ElevenAgents also provides immutable versions, branches, and percentage-based traffic deployment.
What node types are available?
Current documentation describes subagent, dispatch-tool, agent-transfer, transfer-to-number, and end nodes, connected by conditional or unconditional edges.
How do you test a workflow?
Use full simulation tests, next-reply tests, and tool-call tests. Run important cases repeatedly because agent outputs and LLM-based routing can vary.
How much do Agent Workflows cost?
The builder is included from the Free plan. Paid ElevenAgents plans range from $6 to $990 per month before Enterprise; LLM, telephony, extra minutes, burst usage, and text-message charges can add to the total.
Do I have to tell callers they are speaking with AI?
Yes. ElevenLabs' terms guidance requires clear notice immediately before the interaction that the user is dealing with AI and that conversations are recorded and may be shared with ElevenLabs and third-party model providers.
Can a workflow replace human support?
It can automate bounded, well-tested paths, but organizations should retain human escalation for uncertainty, complaints, emergencies, sensitive users, and consequential actions.
Bottom line
Our Agent Workflows verdict
ElevenLabs Agent Workflows is a strong fit for teams already building on ElevenAgents that need clearer branching, specialized conversation phases, explicit tool execution, and staged releases. Its graph, testing, analytics, and versioning are meaningful operational features. The main risk is mistaking visual structure for deterministic behavior: production quality still depends on narrow permissions, repeated tests, safe fallbacks, cost accounting, disclosure, privacy controls, and human escalation.
Visit Agent Workflows website ↗Similar Agents Tools
View all Agents AI tools →
Anything
Anything - AI agent that ships mobile and web apps without coding
Agent Builder
Agent Builder - OpenAI's visual canvas for creating and versioning multi-agent workflows
Action Agent
Action Agent - Writer's enterprise-ready autonomous AI that works on your behalf
Petri
Petri - Anthropic's open-source agentic system for evaluating LLM safety
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.