Create email-based agents and visible workflows for scheduling, inbox work and connected business tasks, with scoped permissions.
Carly AI
Explore features, practical uses and pricing below.
Carly AI is a connected assistant and agent platform for email, scheduling and business workflows. Users can describe work in an email and create agents that operate across connected systems. It suits professionals coordinating calendars and correspondence, as well as teams that want recurring operational work represented as an inspectable process.
A custom agent can have its own name, email, integrations, permissions and files. This makes it possible to separate an executive assistant from a more narrowly scoped agent for a particular business process. The important setup decision is which information and actions each agent actually needs.
Carly describes scheduling, follow-ups, research and CRM work alongside connected Google and Microsoft email or calendars. Its workflow product turns a conversational request into steps that users can inspect and change. The documented building blocks include triggers, tool calls, branches, filters, loops, waits and sub-workflows.
This supports both predictable automation and steps that need an AI interpretation. For example, moving a message from a known sender into a tracker is different from reading an unfamiliar request and deciding how to classify it. Keeping that distinction visible helps explain when an AI step is necessary and which part of the process needs review.
The MCP integration exposes Carly tools for email, calendars, CRM, workflows, files and notes to compatible assistant clients. It is another way to use connected capabilities, rather than proof that an external assistant automatically has access to every account.
Review the account connections and agent permissions before allowing an action such as sending a message or changing a calendar. Shared agents and organization features can support team work, but responsibilities still need to be clear: who maintains the workflow, who sees its data and who checks a mistaken result.
A consultant could ask Carly to monitor a designated inbox for meeting requests, extract the proposed dates and prepare the next scheduling step. The workflow could filter irrelevant messages, check the connected calendar and branch when the request is incomplete or conflicts with an existing event.
Before enabling routine use, try requests containing several time zones, a rescheduled meeting and a forwarded thread with outdated details. Inspect the resulting steps and calendar changes. A separate follow-up workflow could add a reminder after the meeting. These are suggested evaluation tasks; the suitability of an automated action depends on the configured tools and the information available.
Ambiguous messages, stale contact information and disconnected accounts can affect an agent's work. An understandable workflow does not guarantee that every extracted date or drafted reply is correct. Keep consequential steps within the permissions you intend and review exceptions before broadening the process.
The pricing page lists a free option with MCP access, ten workflows, a multi-calendar view and task management. Personal is displayed at $35 per month with 200 credits; team and organization options have different arrangements. The vendor distinguishes free deterministic workflows from credit-consuming AI work. Confirm actual credit use and separate communications add-ons before budgeting a recurring agent.