Define
Draft an entity type or a reusable mixin.
AI-first catalog engine
A product catalog that remembers every change.
Attricat is AI first. Agents can look through your catalog and draft edits, and a person signs off before anything is saved. Every record also points to the schema revision it was created with, so changing the model never reshapes old data behind your back.
Example view of the Attricat entity explorer
Built into the core
Schema revisions that never changeValues inherited by contextSearch that follows relationshipsA change log with permissions
Versioned schemas
Publish a new schema. Old records stay put.
A blueprint describes an entity type: its attributes, mixins, validation rules, relationships, files, and views. Publishing freezes that revision for good. Each entity stays on the revision it was created with and only moves when someone migrates it.
Publish
Freeze the draft as a revision. New entities use it.
Migrate
Look over the outdated entities, then move them to the new revision when you’re ready.
What ships today
The parts you need to run a catalog day to day.
Agents work in the same workspace as your team, and anything they want to change goes through an approval. The features below cover the rest of the daily work around structured catalog data. Attricat isn’t a full PIM or DAM suite, and it doesn’t try to be.
Values that change by market
Model markets, channels, stores, locations, languages, or any other hierarchy. A value set higher up is inherited below until someone overrides it.
Edits respect permissionsSearch across relationships
Search a record’s values or the records linked to it. Filter by type, sort by related values, and share the result as a URL.
Up to three relationship hopsMembers, roles, and an audit log
Invite people, give them built-in or custom roles, and issue personal tokens. The activity log shows who changed what.
Roles and activity logChecks and small automations
Versioned rules flag missing, stale, or tagged data. Workflows run by hand, on an event, or on a UTC schedule, and each does one narrow job.
Findings, retries, audit trailGet data out
Export channels and publications carry contextual values, relationships, and files, plus signals about data health.
S3-compatible file storageAPI, CLI, and extensions
Script against the HTTP API or the JSON-first CLI. Extension packages can add workspace behavior and UI, and you review their permissions before anything runs.
Installed off. You turn them on.Entity explorer
Current and historical data in one view.
Pick a blueprint revision and a context. Values resolve for that context, and each entity shows its files, relationships, revision, and history. Records on an old revision get marked as outdated. Attricat won’t quietly rewrite them.
Search linked records
Query one field, or follow relationships with an explicit path.
Filters that know the type
Text: equals, contains, starts with. Numbers, dates, and times: equals and ranges. Booleans: equals.
Share the exact view
Attribute and relationship filters live in the URL, so a link opens the same view for whoever clicks it.
Governance
Keep control without slowing down every edit.
Permissions, publication state, quality findings, and the activity log all live in the same model, so there’s no extra tool to bolt on.
Workspace roles
Decide who can do what, and give each role its own explorer shortcuts.
Rules and findings
Run checks on demand, as a dry run, on a schedule, or when an event fires.
Versioned workflows
Small actions on a single entity. Failed runs retry, dead letters get parked, and everything leaves an audit trail.
Publication control
Editing a published entity can pull its channel approval. You can exempt trusted roles.
Optional AI agents
The agent asks before it writes.
Connect an OpenAI-compatible provider you trust and agents can look around a workspace and propose edits. Read-only tools may run on their own. Anything that changes data waits for a person to approve it, and that approval is stored.
- API keys stay on the server
- One approval covers one proposed action
- Conversations and actions stay on the record
Who uses it
Three kinds of people work in Attricat.
Catalog builders
Design blueprints and mixins, write validation, set up views, and install extensions.
TOML blueprintsExtensionsMigrations
Catalog operators
Keep entities accurate day to day. That means context overrides, quality checks, publishing, access, and audits.
ExplorerRulesWorkflows
Developers & integrators
Talk to the Rust HTTP API or the JSON-first CLI, and hook into files, events, and extension points.
APICLIS3-compatible files
Try it
See how your catalog would look in Attricat.
The guides go one task at a time through blueprints, contexts, the explorer, agents, and workspace admin. Or skip ahead and read the source on GitHub.
Questions about fitting Attricat to your catalog? Write to hello@attricat.com.