Carly AI: Email-Native Executive Assistant, Mailbox Delegation, and Pricing Explained
- 1 hour ago
- 5 min read
Carly AI, formerly called CalBot, is an AI executive assistant built around email as the primary interface rather than a chat window or a separate app: a person creates named agents by emailing Carly directly, each agent gets its own email address and a set of plain-English instructions, and triggering an action is as simple as CCing that agent on a thread the way a colleague would CC a human assistant. Connections span more than 200 integrations — Gmail, Outlook, Google and Microsoft calendars, Salesforce, HubSpot, Asana, ClickUp, Slack, Zoom, and Teams — so an agent reads and writes data across whatever tools the underlying task touches.
The mechanism that actually separates Carly from a general AI assistant's email features is mailbox delegation: given a mailbox owner's address, an agent can read that mailbox, browse its folders, move messages, and send on its behalf — a capability multiple independent comparisons note isn't available natively in ChatGPT's or Claude's email connectors, both of which require a person present in the conversation to confirm each send. For someone evaluating Carly, the deciding factor is whether autonomous execution across a delegated mailbox and connected tools — completing tasks while the person is away, not drafting them for approval — justifies a dedicated subscription over the email capabilities already bundled into a general AI assistant subscription.
··········
HOW NAMED AGENTS AND MAILBOX DELEGATION ACTUALLY WORK.
Per-agent email addresses, delegated mailbox access, and a cross-assistant interoperability role define the mechanism beyond a chatbot with an inbox connector.
Each Carly agent is a distinct entity with its own email address, its own persistent memory, and its own written instructions — a scheduling agent, a lead-follow-up agent, and an invoice-reminder agent can run independently and simultaneously rather than sharing one general-purpose assistant context. Triggering one requires no separate login or app: forwarding an email to the agent's address, or CCing it on an existing thread, is the entire interaction model, which is also why setup for a new user or a new colleague working with an existing agent requires no onboarding beyond knowing the address to email.
Mailbox delegation extends that model beyond a person's own inbox: granted access to someone else's mailbox, an agent reads incoming mail there, takes action — scheduling, replying, updating a CRM record — and does so under that mailbox's identity rather than its own, which is the specific capability that makes Carly usable as a genuine delegate rather than a personal-only tool. A newer role positions Carly as connective tissue between other AI assistants themselves: documented examples show Carly bridging multi-tenant access that ChatGPT, Claude, or Gemini's own connectors don't support natively — connecting multiple Microsoft Teams tenants to a single assistant, for instance, where those other tools are limited to one tenant or read-only access.
........
Component | Mechanism | Function |
|---|---|---|
Named agents | Each with a distinct email address, instructions, memory | Runs multiple independent, specialized assistants simultaneously |
Trigger interface | Email or CC, no separate app required | Matches the interaction model people already use with human assistants |
Mailbox delegation | Reads, browses, and sends from a delegated mailbox | Acts on someone else's inbox under that mailbox's identity |
Integrations | 200+ across email, calendar, CRM, project, and support tools | Lets agents read and write data across the tools a task actually touches |
Workflow builder | Web dashboard with role-based templates | Chains multi-step automations without building them from scratch |
Cross-assistant role | Connects to ChatGPT, Claude, Gemini via MCP/registration | Extends multi-tenant or delegated access those tools lack natively |
........
··········
WHAT "EXECUTES INSTEAD OF DRAFTS" ACTUALLY MEANS, AND ITS REAL FRICTION POINTS.
Autonomous sending is the productivity gain and the governance question at the same time.
The practical difference between Carly and a general assistant's email connector is timing: ChatGPT and Claude can draft, and on paid plans even send from a connected account, but both require the person to be in the conversation confirming each action before it goes out. Carly's agents send while the person is away — a scheduling thread gets resolved, a CRM record gets updated, a reminder goes out, all without anyone opening the inbox in real time. That's the entire productivity case for the tool: work that would otherwise wait for the person to sit down and respond instead completes on its own.
That same autonomy is also where deployment friction shows up. Granting an agent delegated access to a shared or organizational mailbox on the Microsoft side requires a Microsoft Entra administrator to grant org-wide consent, and admins separately choose which actions are available — a genuine IT governance step rather than a one-click connection, precisely because the agent is sending mail and taking actions under someone else's identity rather than its own. Any team considering Carly for shared or role-based mailboxes, rather than a single person's personal inbox, should expect that admin approval step as part of rollout, not an edge case.
··········
HOW CARLY'S PRICING COMPARES TO SPECIALIST AND GENERAL-PURPOSE ALTERNATIVES.
The premium over a general AI subscription buys autonomous execution and multi-agent breadth, not just an inbox connector.
Carly's Personal plan runs $35 per month billed annually and includes unlimited custom agents, unlimited Google and Microsoft account connections, a workflow builder with no cap on automations, and persistent agent memory; Enterprise adds custom branding, admin controls, API access, and dedicated onboarding at custom pricing. That sits above both a narrow scheduling specialist and a general AI subscription, reflecting a broader scope than either.
........
Tool | Model | Starting price |
|---|---|---|
Carly AI | Multi-agent, email-native, autonomous execution | Personal $35/mo (annual); Enterprise custom |
Calendly | Scheduling links, no autonomous email agent | ~$10–12/mo |
ChatGPT Plus / Claude Pro | General assistant, drafts with user confirmation | ~$20/mo |
........
The comparison isn't which tool is cheapest — a scheduling specialist like Calendly does one job well at a much lower price, and a general AI subscription already covers drafting and occasional confirmed sends at no extra cost beyond what most people already pay. Carly's price buys the specific combination of autonomous action, mailbox delegation, and multi-agent specialization none of the cheaper alternatives offer together, which only pays off if a person's actual workload spans exactly that combination rather than one piece of it.
··········
THE DECISION RULE FOR EVALUATING CARLY AI.
Carly earns its subscription in proportion to how much of a person's or team's coordination work is genuinely repetitive, cross-tool, and safe to complete without real-time confirmation — scheduling logistics, routine CRM updates, standard follow-ups — since that's exactly the work autonomous execution replaces. Someone whose email and calendar needs are occasional or who wants to review every outgoing message before it sends gets little benefit from the autonomy that defines Carly's value, and a general AI subscription or a narrow scheduling tool covers that lighter need at a fraction of the cost. Teams considering Carly for shared or role-based mailboxes should plan for the Microsoft Entra admin-consent step as part of the rollout timeline rather than assuming delegated access connects as easily as a personal account does, since that governance step is a real deployment dependency, not an optional configuration detail.
·····
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····