Skip to main content
Open a mission, then the Configuration tab. The settings sit in nine cards. A closed card shows a one-line summary and an icon for saved, unsaved, or needs setup.
The Configuration tab
After you change something, click Save changes in the bar at the bottom, or Discard. On a live mission, active runs use the change from their next step.

Goal

The Goal card
  • Name is what your team calls the mission, up to 200 characters.
  • Goal is what the agent works toward with each person, up to 2,000 characters.
  • Owner is the teammate the mission acts as. It can only use what the owner is allowed to use. Giving it to someone else needs access to manage members, and the new owner needs access to change Missions.

The mission succeeds when

Outcomes say what counts as reaching the goal. Click Add outcome and fill in:
  • Outcome, a short name such as “Call booked”, up to 120 characters
  • Counts when: When they reply lets the agent decide from the conversation. Marked by hand waits for you to mark it on the run.
  • Worth, a dollar value up to $10,000,000. It adds up on the Outcomes tab.
Every mission needs at least one outcome.

Milestones on the way

Milestones are steps toward the goal, such as “Champion found”. The agent uses them to judge progress, and the Outcomes tab shows where runs stall. Click Add milestone. They’re optional.

Who and when

The Who and when card
Start a run when holds the rules that start runs on their own. With no rules, you start runs by hand. Signals and events explains rules. One run per picks who shares a run. With Person, everyone gets their own. With Company, the default, people at the same company share one. People with a personal email, like Gmail, always get their own run.

Strategy

The Strategy card
Brief the agent like a new teammate. Your AI agent’s tone and policies still come from its Instructions. The editor takes up to 50,000 characters. Type @ to mention an action that’s on. Type {{ to insert a contact or company detail, such as {{contact.name}} or {{company.industry}}.

Steps it must always take

These are steps the agent takes in every run, whatever else it decides. Click Add step, then set When, like “Day 3”, and the Step, like “Send a WhatsApp message to their manager”. You can add up to 50.

Channels and actions

Turn on what this mission can use. An action that isn’t ready shows Needs setup until you connect what it needs.
The Channels and actions card

Reach out

Email to an address that unsubscribed or bounced is blocked.

Your team and data

  • Ask me when stuck. The agent asks you a question and waits. After 3 days with no answer, it decides itself. You can edit when it should ask.
  • Create a task. Opens a Linear issue for a teammate when something needs a person, like a LinkedIn connection. Closing the issue wakes the run. Pick a Linear team and who gets tasks, either one teammate or rules you write. Needs Linear connected in Integrations.
  • Edit a contact. Updates a contact’s details.
  • Add a contact to the run. Brings another person in, such as a manager.
Read and search, and the AI agent's Actions
Open Read and search to choose what the agent can look up:
  • Research a person searches the web for their recent posts, talks, and news. Results last 7 days.
  • Read a conversation and Search conversations
  • Read a ticket and Search tickets
  • Find contacts at a company
  • List WhatsApp templates
The agent can always search your Data Sources and read links.

Your AI agent’s Actions

Under the built-in actions, the card lists your AI agent’s own Actions that missions support:
  • Custom actions that call your server
  • Shopify: show an order or product, get metafields, tag an order, create a return
  • Stripe: get a subscription or invoices, find payments, refund a payment
  • Calendly and Cal.com available times
  • Web search
An Action only shows here when it’s enabled and not limited to Procedures. To keep an Action out of every mission, open it on the Actions page and turn off Missions under its channels.

Approvals

The Approvals card

Mode

Run actions without approval is off by default, so every message, call, and change waits for you. Turn it on and emails, WhatsApp messages, calls, and in-app messages go out on their own. They still follow your guardrails. Some actions always wait, even with this on:
  • Edit a contact
  • Add a contact to the run
  • Your AI agent’s Actions that change data, like a Stripe refund
Actions that only read data, and Create a task, never wait.

Routing

Send approval requests to picks where approvals go:
  • No one keeps them on the run page. This is the default.
  • Email sends them to the teammate set in Alerts.
  • A Slack channel posts them with Approve, Skip, and Open run buttons.
  • A Slack DM messages one teammate. If Slack fails, it emails instead.
When approvals go to email or Slack, If nobody decides sets what happens after a few days: send it again, or skip it. The default is to send it again after 3 days.

Guardrails

The Guardrails card
Credits per month caps what this mission spends. The default is 1,000. When it’s reached, runs pause until you raise it, then pick up on their own. Credits explains the cost. Outreach limits cap how much this mission sends. Leave a field empty for no limit.
  • Messages per person per day
  • Messages per person per week
  • People per company per week
Send only during set hours holds messages outside the days and hours you pick. It uses the contact’s own time zone when it’s known.
Sending hours and the Never contact list
Never contact lists emails, domains, or phone numbers the mission must skip, up to 1,000. When a guardrail holds a message, the run shows On hold with the reason, such as “Quiet hours” or “Already reached 3 of 3 people at this company this week”. If someone replied in the last 15 minutes, the next message waits too. Every account also has a daily cap across all missions: 1,000 emails, 1,000 WhatsApp messages, and 200 calls.

During a run and when it ends

The During a run and when it ends card
While a run is going holds rules that change a live run, like waking the agent when they start a trial. See Signals and events. Stop or hold the run protects your customers: A held run waits for a teammate to resume it. You can add your own hold and stop rules. Give up if the goal isn’t reached within ends runs after 7 to 90 days. New missions use 60. A run also ends after 12 steps.

Voice

The Voice card
Writing style sets the tone, length, sign-off, and language of this mission’s emails, messages, and call greetings. It wins over your AI agent’s instructions when they differ. Example messages are up to 3 messages you’d write yourself. The agent copies their tone and length, not their content. Exact wording is text the agent sends word for word, like a legal line. Pick whether every email or every call must include it. A message without it can’t be sent.

Alerts

The Alerts card
Email important updates sends a short email when something happens on a run, like a reply or a question for you. It’s off by default, but a run paused by errors always emails.
  • Send emails to is the mission owner by default.
  • Also email adds up to 10 teammates. They get updates, never questions or approvals.
  • Briefing emails go out right away, or once a day as one summary.
Slack messages post to your team’s Slack when something happens on a run. The agent writes each one, and they don’t need approval. Click Add Slack message, pick a channel, a direct message, or a teammate’s App Home, and say when to send it. You can add up to 10. Slack must be connected to the AI agent.