GET/v1/skill
The CRM is dumb.
The agents are smart.
Seven rules, and the whole contract is in them. This is the part of the documentation an agent is meant to read twice — it is short because the product is: the CRM decides nothing, it only remembers, and it refuses to forget.
Quoted from the skill file, with the reason each one is there
The left column is what an agent reads. The right is why we wrote it that way — which an agent does not need, and a person evaluating the product does.
You are the user.
Everything the dashboard can do, you can do. If a capability is not in the skill file, it does not exist.
A dashboard that can do something the API cannot is a promise the product breaks the first time an agent needs that thing at three in the morning.
The timeline is the truth.
Every field is the latest known value derived from it. To know why a record says what it says, read its activities.
Fields drift, get overwritten and lose their reason. An append-only log does not, so the log is what the fields are derived from rather than a report about them.
Every fact has an origin.
Send source and confidence with anything you write. A later reader has to be able to judge your work.
Two agents will disagree about the same company. The one that can be judged — by source, by confidence, by who wrote it — is the one a human can act on.
Nothing is deleted.
Set a status and a reason instead.
Deletion destroys the evidence that a decision was made. A status and a reason keep it, and the thirty-day purge on hidden schema elements is the only exception.
Upsert, never duplicate.
Natural keys make repeated calls safe. Send the same payload twice and the second call changes nothing and writes nothing.
An agent that retries is not a bug. Natural keys mean the retry is free, so nobody has to build a deduplication pass over the top.
Humans correct, agents propose.
Think two records are the same? Say so with a note. Do not merge. A human decides.
Merging two records is a judgement about the world, not about the data. An agent can flag it with a note; a person decides.
The CRM is dumb, you are smart.
No rules engine, no sequences, no sending. You chain tasks to make things happen.
Rules engines and sequence builders are where CRMs go to become unexplainable. Chained tasks do the same work in a form you can read top to bottom.
What this is not
Said plainly, because every one of these is something a CRM is usually expected to do, and an agent that assumes we do it will write code that never runs.
Not an agent
It has no model behind it and makes no decisions. Your agents bring the judgement; this is where they write it down.
Not a sender, and not an SDR
It will never email, message or call anyone for you, and it prospects for nobody. You bring the search, the enrichment and the outreach, and record the results here.
Not a sequence engine
No cadences, no branching automations, no rules to debug. A task that creates its follow-up is the whole mechanism.
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.