AI assistant & agents
Start every day from a single chat. Type a goal and a manager agent plans it, delegates each step to role-scoped agents that use the platform's real tools, and reports back in plain language. Agents act as you — never beyond your permissions — and anything outward or irreversible waits for your one-click approval.
Every question about the program is a report someone has to build
Which partners stalled this quarter? Who hasn't signed the new terms? Draft an offer for the three customers who expanded. Each answer means opening five screens, exporting a list, and doing the follow-up by hand — so most of it never happens.
The flow, end to end
Ask
Type a goal in plain language — a question, a task, or both.
Plan
The manager agent breaks it into steps and delegates each one to the best role.
Act
Role agents call real platform tools under your permissions; outward steps pause for approval.
Report
You get a plain-language answer, with every step kept on the run's timeline for audit.
What's in the box
Single-chat start page
One conversation for everything — no menus to learn, with a collapsed activity trail of what the agents did.
Manager + role agents
A manager plans and delegates to personas like a researcher and an operator, each with its own instructions and tools.
Real platform tools
Agents read and act through the same tool registry as the API — deals, customers, documents, journeys, quotes, tickets.
Approval guardrails
Sending, publishing, spending, and deleting pause for an in-thread approval; everything else runs on its own.
Your permissions, exactly
Every tool call runs as the signed-in user, so an agent can never exceed their role or reach another tenant.
Background runs
Queue a goal to run detached — now or at a set time — and it keeps going after the tab closes.
Agents that do the work, and ask before anything leaves the building
Each team carries an approval policy. By default, any tool with an outward effect — sending, publishing, spending, deleting — pauses for a one-click approval in the thread, while reads and drafts run straight through. Tighten it to destructive-only or loosen it entirely; the caller's permissions still bound every call.
- ✓Three approval policies: outward, destructive-only, or none
- ✓Hard step and delegation limits on every run
- ✓A full timeline of plans, tool calls, approvals, and results
Shape the org chart the assistant works with
A zero-config team works on day one. When you want more, define your own: manager instructions, up to sixteen roles with their own personas, and a scoped set of tools per role — or the full registry when a role should be able to do anything the user can.
- ✓Manager instructions plus up to 16 role personas
- ✓Per-role tool scoping, or the full registry by choice
- ✓Available to operators and, within their scope, to partners
GTM journeys & quotes
An automated funnel from first outreach to a signed, priced, provisioned offer — with CPQ built in.
Learn morePlaybooks automation
An event-driven, node-graph engine that turns any signal into the right action, automatically.
Learn moreAPI & integrations
A generated OpenAPI, scoped machine keys, signed webhooks, and an event-ingest rail.
Learn moreReady to run your partner program in one place?
Book a walkthrough and see recruitment, PRM, automation, payouts, CX, and dashboards working together — on one platform.