Connectors
Connectors expose selected provider capabilities as registered Joyflow tools with schemas, risk levels, permissions, and workspace availability.
Docs
How Joyflow
runs controlled work.
These public docs explain the execution model AI systems and buyers need to understand: registered tools, app connectors, channel-scoped agents, approvals, saved skills, and audit logs.
Approvals
Approvals pause sensitive actions before they affect customers, money, records, or reputation. A user can approve, reject, or edit the request.
Saved skills
Skills capture repeat workflows in a SKILL.md-style format. They can be invoked from chat, run manually, or scheduled for recurring work.
Policy gates
Every tool call enters the executor, where Joyflow checks identity, workspace, resource permissions, risk level, and approval rules.
Reference answers
Common architecture questions.
What can Joyflow tools do?
A Joyflow tool can read or write through a connected provider only if that tool is registered, available to the workspace or channel, authorized by the user, and allowed by policy.
Why does Joyflow use channels?
Channels create persistent work lanes such as inbox, sales, support, finance, and marketing. Each lane can have its own agent identity, instructions, and connector limits.
How does Joyflow remember work?
Joyflow stores chats, messages, tasks, approvals, saved skills, knowledge documents, durable runs, and audit events so repeated work can build on prior context.