Actioner's customer graph is structured — companies, contacts, deals, every interaction, captured automatically. Artifacts extend it with everything that isn't: usage snapshots, support exports, contract notes, any data in any shape — attached to the account, stored on your machine, readable by every play.
Every structured system eventually meets data that doesn't fit its fields. The usual answers — custom objects, another column, another admin project — arrive months late and fit poorly.
The traditional fix is an integration per source: built, maintained, broken, rebuilt. Most useful data never clears the bar — it simply stays out of reach.
So the real story lives in exports, spreadsheets, and one-off docs — true, useful, and invisible to every tool that could have acted on it.
A play reaches a tool through its MCP connector — PostHog, Amplitude, wherever the data lives — and saves the result to the account. Schedule it, or let another play trigger it on demand.
"Pull this quarter's usage for Meridian and save it as a snapshot."
Working through something in Claude — an analysis, a research pass, a summary worth keeping? Save the result to the account, and it becomes part of the memory every play can use.
"Save this competitive summary to the Acme account."
Drop in the file that never had a home: the NPS export, the contract notes, the spreadsheet from finance. Structured or not — no schema to fit, no importer to fight.
"Here are the support tickets from last quarter."
The moment an artifact lands on an account, every play can draw on it. Churn analysis reads the usage snapshot. Meeting prep reads the support summary. Account reviews read the contract notes. Nothing to configure — the data is simply there, where the account is.
And when a play cites an artifact, the claim carries its source and date — the same receipts discipline as every other insight in Actioner.
Artifacts are stored locally with the rest of your graph — not on Actioner's servers. Same local-first architecture, no exceptions.
Claude reaches artifacts over the same MCP connection as the rest of your data — scoped to your access, seeing only what you can see.
Add what helps, update what changed, delete what shouldn't be there. The memory is yours to shape — plays adapt to whatever it holds.
See artifacts at work: Churn Analysis uses them to add usage data from PostHog or Amplitude to its relationship signals — the difference between churn analysis that's relationship-aware and one that sees the full picture.
"If it says something about the customer, it belongs in the memory."
Available today · Any data, structured or not · Stored on your machine