Capabilities

An agent is a Definition — the document you write — plus Capabilities: the functionality you switch on. A capability is a bundle: the runtime machinery that does the work, the settings you configure, any account connections it needs, and what the agent itself learns to do while it's active. You don't assemble those parts; you enable the capability, and the platform wires all of them.

Capabilities are configured, not authored. Each one is described here by what the agent can do when it's on, what it needs from you, and where its settings live in the console. The Capabilities page in the console shows the same roster with each capability's live status.

The capability lifecycle

Every capability moves through the same stages:

1. Enable. Some capabilities are explicit switches; others turn on when their content or configuration exists (adding documents enables Knowledge; setting availability enables Scheduling).
2. Configure. Supply the settings the capability needs — hours, fields, templates. Configuration is structured (forms), never prompt text.
3. Connect. Capabilities that act through an external account (Google Calendar, Slack) need a connection. Connections belong to each agent individually and their status is shown per item at Settings · Tools & connections.
4. Ready. The capability's machinery runs, and Playbooks can use it. When something is missing, the console says what — a capability with incomplete configuration is not offered to the agent, so it fails visibly in setup rather than silently in conversation.

Playbooks and capabilities meet in the Definition: a Playbook that handles booking relies on Scheduling; a product-questions Playbook relies on Knowledge. The agent is offered a capability's actions when the active Playbook calls for them and the capability is ready.

Scheduling

The agent can book, reschedule, and cancel calls between people and your representatives — the humans who take the calls. It proposes real availability, confirms exact times in the person's timezone, issues calendar invites, and handles "when's my call again?" later. Behind the scene, the platform also runs the machinery you'd otherwise build: reminders, day-of completion, double-booking guards, and a deterministic backstop that shows times even when the model falters.

Configuration: availability (business hours, blackout dates, call duration, same-day policy) at Settings · Scheduling, and the representative directory at Settings · Representatives — names, titles, phone numbers, and optionally a Google Calendar connection per representative. Representatives are part of scheduling's configuration: they're who the booked calls are with, and each Person can be assigned one. With no representatives configured, booking still works — calls are scheduled with "someone from our team" — but the capability reports it as needing attention, and per-representative features (the rep card, tap-to-call, calendar sync, notification email) stay off.

Guide: Scheduling Calls.

Knowledge

The agent grounds its answers in what you give it: documents (policy pages, pricing, FAQs — uploaded or imported from your website) and products (the structured catalog it presents as rich cards and records interest against). On every turn the platform retrieves the material most relevant to what the person said; the agent answers from it rather than improvising, cites public sources where appropriate, and never reveals internal file names.

Knowledge is on when content is present. Documents and products are managed in the agent's Knowledge area.

Guide: Knowledge & Products.

People

The agent remembers the human it's talking to. One Person record per person, per agent — the built-in CRM; no external system to wire up. The record is on the agent's desk every turn: contact details, the People fields you define ("target market", "plan tier"), product interest reconciled against your catalog, funnel status, an assigned representative, and a memory journal of what the agent has learned.

Configuration: the field schema at Settings · People fields, and one decision — whether the agent may update People with what it learns. Turned on, it saves facts as they come up naturally (never by interrogating); turned off, it reads but never writes. Person records themselves live on the People page, which is also where day-to-day work happens: import, segments, outreach actions.

Guide: People.

Outreach

The agent can start conversations instead of only answering them: a reviewed opening Message Template goes out on a supported channel (RCS today), automatic follow-ups nudge people who don't reply — within quiet hours you set — and the conversation then proceeds exactly like an inbound one, starting in the Playbook chosen for the send (or the one you've marked default for outreach).

Outreach has two faces. Its configuration (greeting template, follow-up steps and cadence, quiet hours) lives at Settings · Outreach. Its operations — actually reaching out to a person, pausing a follow-up cadence, minting shareable outreach links, inviting testers — happen on the People page and the Operations section, day to day.

Guide: Reaching Out to People.

Notifications & connected accounts

The agent can alert your team when something needs attention — a booked call, an escalation, an event you care about — through the service you choose: Slack, email. The agent's vocabulary never changes with the service; you pick where alerts land, and you can swap the service without touching any Playbook.

Connections and per-item status (connected, expired, needs reconnection) live at Settings · Tools & connections. Secret values — API keys, webhook URLs, auth headers — are write-only: stored encrypted, shown as set with a Replace action, and never included in an exported document.

Guide: Capabilities in the console.

Custom integrations

Every business has systems of its own. A custom integration gives the agent an ability backed by your HTTP endpoint. To the agent it's simply something it can do mid-conversation; you implement what happens.

You define one by briefing the agent — a name, a description of when to use it (written the way you'd brief a new teammate), and the inputs it should gather from the conversation — plus the endpoint the platform calls: method, URL, and an optional write-only authentication header. When the agent decides the ability applies, the platform calls your service with the inputs, hands the response back, and the agent continues the turn.

Defined at Settings · Custom integrations.

Guide: Capabilities in the console.

Capabilities and the portable document

An exported agent carries capability configuration in its capabilities layer — scheduling settings and representatives, people fields, products, outreach settings, custom integration contracts. It never carries connections, secrets, or data (documents, Person records): those are per-agent and re-established after an import.