← All agent and harness briefs
Agent runtimeautonomousopsgtmvoicedocumentedunattended capable

Operational brief

Carly

Email- and calendar-native operations agents with persistent memory, scheduled tasks, and connected business workflows. Vendor guides also document inbound phone work and scheduled briefing calls.

No product visual on file

Choose it when

You want a persistent operations assistant that can work through your existing email and calendar, including recurring follow-up and scheduling. You also want vendor-documented phone access to connected tools or scheduled briefing calls with follow-up questions and an emailed rundown if missed.

Skip it when

You need a self-hosted agent, formal public assurance evidence before connecting sensitive accounts, or independently verified enforcement of approval before every outgoing action, rather than configurable per-action rules. For phone work, skip if you need desk-verified caller authentication or guaranteed call timing/delivery.

01 / Classification

Autonomy evidence

This is our classification, not a vendor badge. The record says what the product normally does. The gaps stay visible.

How we classify it
Unattended scheduler/heartbeat with write-tools; failure has a real owner.
Runs unattended
Yes — that capability is documented in this brief.
Trigger
Paid custom agents support scheduled tasks, recurring or one-off, that run automatically; work can also arrive by email/CC. This is optional configured delegation, not a claim that every free MCP call is autonomous. Vendor guides document scheduled briefing/reminder calls and incoming owner or configured trusted-caller requests; other callers reach the receptionist.
State
Paid agents have persistent memory for contacts, preferences and past interactions; the controls article describes standing instructions read at each task and a queue for held actions. The phone guide also describes logging calls, transcripts and texts to the person's record; retention and completeness aren't independently verified.
Tools
Agents can send email, book meetings and update CRM records through connected accounts. Actions outside configured require-approval rules can run without a human click. Vendor phone guides also describe owner calls accessing connected email/calendar, sending texts and placing calls, plus scheduled briefings with spoken follow-ups and email of a missed briefing.
Human control
Configure require-approval rules per action and assign a job owner. The team guide assigns instructions, approval rules and queue oversight to an Owner and owned sends to teammates. That owner must request/review the recommended sent/booked/held/failed digest and resolve exceptions; it isn't a guaranteed recovery service.

02 / Operations

Control reality

Useful agents eventually touch something real. This is the operational picture we have, not a promise hidden in a sales page.

Where it runs
cloud
Data boundary
Not recorded yet.
Identity & permissions
Not recorded yet.
What it can write
email and calendar actions · CRM records · scheduled workflows · connected application data · scheduled phone calls and missed-briefing email · call logs and transcripts
Surfaces
Email and calendar · SMS · Slack · Dashboard · MCP · Phone (owner/trusted-caller access and receptionist)
Export / lock-in
Not recorded yet.
Operational limits
Documentation-only, not a connected-account trial. Require-approval rules are configurable per action; external email can run without approval. A failed-work digest is a setup recommendation, not guaranteed built-in alerting or automatic recovery. The job owner must configure it, review held/failed work, and resolve failures. Phone timing, delivery and caller authentication are not runtime-tested; confirm phone-workflow entitlement, region availability and charges separately. The separate developer CLI/MCP manages booking pages, not hosted email/workflows/CRM or a local general-agent runtime. Current pricing explicitly lists business SMS, calling and the AI receptionist as separately billed add-ons for Organization teams, with book-a-call setup—not free or automatically included in Personal/custom-agent base seats. Exact scheduled owner-briefing entitlement, geography and call charges still aren't established here.
Watch for
Freemium isn't free custom-agent access: the free tier lists MCP, workflows, calendar view and Task Manager; paid plans list custom agents, automatic scheduled tasks and memory. Set approval rules before enabling sends, name the job owner, and request a digest covering sent, booked, held and failed work. The phone guide says carrier verification gates owner access; that's a vendor claim, not spoof-proof authentication established here. It also describes trusted callers with only allowed access, a blocklist, other callers reaching the receptionist and call/transcript/text logging. Review caller permissions and logged data before use.

03 / Receipts

Evidence ledger

Here is what this brief is built on. “Documented” means we found a source. “Tested” means this desk actually ran it. Those are different things.

Evidence status
documented
Source set
8 primary sources
Evidence review
2026-10-21

Corrections

  • 2026-10-14 · Added the previously missing control/owner bundle: per-action require-approval rules, held queue and named job ownership. Qualified Freemium access and the recommended daily failed-work digest; neither universal approval-before-send nor automatic failure recovery is established.
  • 2026-10-21 · Expanded documented scope using the October 20 phone guides: owner/trusted-caller access, scheduled briefing calls, follow-up questions, missed-briefing email and call records. Caller verification is a vendor claim, not desk-audited authentication or a delivery guarantee. The separate booking-page CLI/MCP doesn't expand this hosted agent's runtime. Release fields unchanged; not runtime-tested.

04 / Change log

Related coverage

The brief tells you what the product is. Payloads track the changes that may alter the decision.

No linked change coverage on file yet.