Agentic AI for real operations
Put AI agents to work in real business processes.
Combine AI agents, business rules, software tools, and human decisions. Give AI room to adapt without giving up control.
Use the visual editor, AI-assisted creation, or the JSON DSL.
Validate and release through the dashboard, MCP, CLI, API, or CI.
Observe runs, handle approvals, compare outcomes, and release safely.
Set permissions, audit every decision, and keep autonomy within policy.
See it in practice
Eight workflows, simple to advanced. Try one, then another.
The panel on the right is the real, embeddable Chatterfly widget, not a mockup. Each workflow adds one more agentic idea, and each remembers what you told the last one, so nothing is asked twice.
Start here
For anyone. Simple first, then the agent takes the wheel.
One agent, four answers
Four button clicks. An agent reasons over them and gives you three honest places an AI agent would earn its keep first.
Three ranked starting points and a first step for this week.
- 01Pick the time sink
- 02Say what hurts
- 03The agent reasons
- 04Three starting points
By function
Legal and finance, sales and marketing, product and engineering.
By industry
Healthcare practices, retail and e-commerce.
Deploying soon
This workflow is written and validated -- its live deployment is on the way.
Drops into React, Next.js, plain HTML, WordPress, Shopify, or any page that can render a web component -- see the embedding guide.
For leaders and operators
Adopt agentic AI one controlled process at a time.
Start with a process your team already understands. Keep human approval where it matters, measure every run, and expand autonomy only when the evidence supports it.
Build your first workflowFor CTOs and developers
Ship agentic workflows with MCP, CLI, and CI.
Let coding agents author the JSON DSL through MCP. Validate and deploy from the CLI, version definitions in git, and integrate through the API or embeddable widget.
Open the Quickstart docsWhy Chatterfly
Agency without operational guesswork.
Move from experiments to operations
Turn promising AI prototypes into versioned workflows your organization can actually run and improve.
Set the boundaries of autonomy
Choose where agents can reason and act, where rules must decide, and where a person must approve.
Use one process everywhere
Run the same workflow across web, voice, phone, messaging, hosted pages, and your own applications.
Know what happened and why
Inspect every run, state change, model response, tool call, human handoff, and final outcome.
Build, deploy, operate, govern
One platform from idea to governed production.
Build visually or as code. Deploy through the dashboard or your pipeline. Operate from a shared run history. Govern every action with explicit policy and evidence.
Build
Use the visual editor, AI-assisted creation, or the JSON DSL.
Deploy
Validate and release through the dashboard, MCP, CLI, API, or CI.
Operate
Observe runs, handle approvals, compare outcomes, and release safely.
Govern
Set permissions, audit every decision, and keep autonomy within policy.
Platform capabilities
The parts you need to run complete workflows.
Orchestration
Agents, tools, conditions, loops, code, retries, and schedules
Human participation
Chat, forms, files, approvals, deliberation, and team inboxes
Every surface
Widget, hosted pages, live voice, phone, and messaging channels
Developer platform
MCP, CLI, REST API, JSON DSL, validation, and CI deployments
Production operations
Versioned releases, run timelines, scoped state, and audit trails
Connected intelligence
Leading models, knowledge bases, MCP servers, and business systems