Agents
AI teammates that answer inbound texts and email on the lines you assign them — with an approval queue until you trust them to fly solo.
An Agent is an AI teammate your organization configures once and attaches to specific lines — a phone number, an email address, or several of each. When a message arrives on one of those lines, the agent reads the conversation, looks up whatever context it needs, and handles the reply.
Typical agents:
- a receptionist that answers common questions and routes the rest to you
- a scheduler that handles "can we do Tuesday instead?" back-and-forth
- a campaign responder that fields replies to an outbound text campaign
- a support agent answering from your own knowledge
Agents belong to your organization
Agents are org resources: owners and admins create and manage them, and their conversations are visible to teammates like any other thread. They occupy no seat and never count toward your member count.
The channel whitelist is the boundary
An agent only ever acts on the lines you explicitly attach. Attaching a line does two things at once: the agent starts listening to new inbound messages on it, and gains permission to send from it. It has no access to any other line, and a line can only be held by one agent at a time.
Autonomy modes
Every agent runs in one of three modes:
| Mode | What happens |
|---|---|
| Draft-first (default) | The agent writes a reply and parks it in the approval queue. Nothing is sent until a person approves, edits, or rejects it. |
| Supervised | The agent sends on its own, but hands anything it isn't confident about to a human instead of guessing. |
| Automatic | The agent sends everything it can on its own and escalates only when it is stuck (guardrails still apply). |
New agents start in draft-first mode on purpose — watching the approval queue for a few days is how you calibrate the instructions before letting it send freely.
Guardrails
Agents are governed, not just prompted. Independent of the AI, every turn passes a hard gate:
- Human right-of-way. If a teammate replies to a conversation first, the agent stands down. Anyone with access to a thread can also take the conversation over — the sparkles button in the thread's header shows which agent covers the line, with Take over this conversation / Resume one click away. When a draft-first agent has a reply waiting on that conversation, the same menu shows it inline so you can approve or reject without leaving the thread.
- Machine-mail immunity. Agents never reply to senders CommSync recognizes as automated systems, promotions, or spam — so your agent can't get into a loop with someone else's robot.
- Opt-outs are permanent. A texted
STOP(and friends) puts the sender on the agent's opt-out list immediately, before any AI runs. - Rate limits. Per-contact daily caps and a consecutive-reply limit stop runaway conversations; quiet hours keep the agent silent overnight if you set them. A message received during quiet hours is skipped, not saved for an automatic morning reply; it stays in the inbox for your team.
- Fresh one-to-one messages only. Agents do not reply to group conversations, and a trigger more than one hour old is ignored so a delayed sync or history import cannot produce a surprise reply.
- Budget. Agent replies draw from the same organization AI budget as every other CommSync AI feature — when it's exhausted, agents stop.
Teach it your business
Instructions tell an agent how to behave; knowledge tells it what's true. On the agent's Knowledge tab you can add two kinds of sources, shared across your whole workspace:
- Documents — upload a PDF, Word file, spreadsheet, or plain text (price lists, policies, FAQs, service menus). Up to 15 MB each.
- Your website — give it your site's address and CommSync reads the public pages it links to, so the agent can answer from the same information your customers see.
Once a source shows Ready, tick the checkbox to attach it to the agent. From then on the agent looks facts up in its attached sources before answering — and when it can't find an answer there, it escalates to a human instead of guessing. Each agent only answers from the sources you attach to it, so a billing agent and a support agent can work from different material.
Keep it current
Agents answer from the version you added. When a document changes, remove the old source and upload the new one; for websites, delete and re-add the site to re-read it.
Connect your tools
Beyond knowledge, agents can use external tools — services your business already runs, like a project tracker, calendar, or ticketing system, connected over MCP. Connections belong to your workspace, not to one agent: set a server up once on the MCP page (also under Settings → MCP), then assign it to as many agents as should carry it. Ask can use workspace connections too.
- Browse the full catalogue at Agents → Catalog — every verified server in a card gallery with categories, search, and pagination — and connect one in a click, or add a custom server your own team runs (with an access token if it needs one — stored encrypted, never shown again). Need two accounts on the same service? Connect the server twice.
- Enable only the tools you want. Connecting a server enables nothing by itself: open it, and CommSync lists every tool the server offers so you can tick exactly the ones to enable. Anything unticked stays invisible to every agent, even if the server adds new tools later.
- Assign connections to agents. Each agent's MCP tab picks which workspace connections it carries — a receptionist and a dispatcher can use different tools from the same workspace.
Two things stay true no matter what you connect:
- The agent can never message anyone through an external tool. Tools let it look things up or act on other systems; every reply to a contact still goes through its normal reply pipeline — including the approval queue in draft-first mode.
- Every external tool call is recorded on the turn's Activity trace, so you can always see what the agent looked up and when.
Signing in to a server
Catalog entries marked Sign-in connect through the vendor's own sign-in screen: click Sign in, approve access on their site, and you land right back in CommSync with the tool picker open. CommSync never sees your password — the vendor issues it a scoped access key, which is stored encrypted and can be revoked any time with Sign out (or by disconnecting the server). Every server in the catalog is verified connectable — if it's listed, it works.
Who can use agents?
Owners and admins always have full access. Whether members can see and configure agents (and whether Ask uses MCP tools for them) is a single workspace toggle in Settings → Team → AI agents — off by default.
Every turn is auditable
Each time an agent acts (or deliberately declines to), the turn is recorded on the agent's Activity tab: the message that triggered it, what the agent looked up, what it drafted or sent, and why skipped turns were skipped. If a reply ever surprises you, the answer to "why did it say that?" is one click away.
Your daily brief keeps you posted too: on days your agents were active, the brief includes a short stat line — how many conversations they handled and how many drafts are waiting in the approval queue. Quiet days add nothing.
After a successful send, the agent also keeps a short private memory for that contact: durable preferences, commitments, and open items rather than a verbatim transcript. It uses that summary in later conversations with the same contact.
Plan limits
Agents are a plan feature with two limits: how many agents your organization can have, and how many turns they can use each month. Both are pooled across the whole organization, not per seat.
| Plan | Agents | Agent turns / month |
|---|---|---|
| Starter | 0 | 0 |
| Pro | 1 | 500 |
| Power | 5 | 5,000 |
A turn is counted each time an agent actually engages the AI to handle a message. The guardrail checks that happen before that — skipped senders, quiet hours, opt-outs, human right-of-way — are free; only turns that reach the model count, whether they end in a sent reply, a parked draft, or an escalation.
When your organization reaches its monthly turn cap, agents simply stop responding until the month resets or you upgrade the plan. Nothing is lost: every message still lands in the inbox as usual for your team to handle. The monthly counter resets automatically at the start of each calendar month.
The Agents page shows owners and admins where the organization stands: a usage meter at the top counts turns used this month against the plan's cap, and each agent's card carries its outcomes for the last 30 days — replies sent, escalations, and drafts waiting for approval.
Need more?
Higher agent or turn limits are available as an override on any plan — contact us and we'll set it up for your organization.
Setting up an agent
Create it. Go to Agents in the sidebar, hit New agent, and describe the job in plain language: who it is, what it should do, what it must never do, the tone to use.
Attach lines. Pick the phone numbers and email addresses it should handle. Attach = read + reply on that line, nothing else.
Set guardrails. Daily caps, quiet hours, and how long to wait before replying (so rapid-fire texts get one considered answer, not three).
Add knowledge (optional). Upload documents or point it at your website on the Knowledge tab so it answers from facts, not vibes.
Activate in draft-first mode. Review its drafts in the approval queue. Approve, edit, or reject — then graduate it to supervised or automatic when the drafts are consistently right.
Write instructions like a job description
The best agent instructions read like onboarding notes for a new hire: what the business does, what a good answer looks like, what to hand to a human. Concrete examples beat abstract rules.