Free
$0per workspace
Enough to run real outbound and decide whether this is where your agents should write.
- Records
- 2,000
- Activities
- 40,000
- Agents
- Unlimited
- API calls
- Unlimited
GET/v1/usage
Priced on what you keep
Not on how hard your agents work. One price, per workspace — never per seat. Each tier holds five times the one before it, and every tier allows twenty activities per record, because that is the shape of what an agent writes.
Pro
$49per workspace, per month
A team running outbound, with its agents writing everything they do back here.
- Records
- 10,000
- Activities
- 200,000
- Agents
- Unlimited
- API calls
- Unlimited
Scale
$199per workspace, per month
For a workspace that has become the record for a whole book of business.
- Records
- 50,000
- Activities
- 1,000,000
- Agents
- Unlimited
- API calls
- Unlimited
Enterprise
Talk to ussized with you
One workspace is one database, so past this the ceiling is a conversation rather than a number on a page.
- Records
- 250,000
- Activities
- 5,000,000
- Agents
- Unlimited
- API calls
- Unlimited
A record is any row in any object — companies, people, deals, tasks and anything your agents invented. So a tier holding 10,000 records is not 10,000 companies; it is closer to a couple of thousand accounts with their contacts, deals and work.
Never counted
An AI-native CRM cannot bill by the thing software does, or the cheapest design is always the one with the least software in it. So the meter is on storage, and these are not on it at all.
Agents
Connect as many as you like. The product is worse if a second agent costs money.
API calls
There is no request quota and no per-call price. Poll as often as you need.
Seats
Everyone who should watch the work can watch it.
A new workspace starts free, and starts unverified
An agent can create one with no key and no human at all, so an unclaimed workspace holds 200 records until a person takes it over. Claim it and it is on Free. Upgrade from inside the workspace, whenever the caps start to bite — the error always names the cap it hit.
More time to build
One skill file, one OpenAPI document, one MCP server generated from it. No SDK to learn and no tool logic written by hand.