Data Sharing
Share databases with other Crow users. Three modes give you control over how much data travels and who can access it.
Sharing Modes
| Mode | What Happens | Best For |
|---|---|---|
| Clone | Full copy of the database is sent to the recipient. They get an independent snapshot. | One-time handoffs, small datasets, offline access |
| Federated Read | Recipient queries your database remotely through the gateway proxy. No data is copied. | Large datasets, live data, controlled access |
| Subscription | Recipient receives ongoing updates as you modify the source database. | Collaborative projects, shared reference data |
Clone
A clone sends a complete copy of a database file to a contact. The recipient gets their own independent copy — changes they make don't affect your original.
"Clone my county-data database and share it with Robin"The AI uses crow_share with share_type: "database" and mode: "clone". The database file transfers through Hypercore, the same P2P channel used for memory sharing.
When to Clone
- The dataset is small enough to transfer comfortably (under ~500 MB)
- The recipient needs offline access
- You want to hand off a dataset without maintaining a connection
Federated Read
Federated read gives a contact permission to run read-only queries against your database through the gateway proxy. No data is copied — queries execute on your machine and only results travel over the network.
"Give Robin read access to my tax-filings database"The recipient can then query your database from their own AI or Data Dashboard. Requests are authenticated with bearer tokens and subject to the same safety model as local queries.
When to Federate
- The dataset is large and cloning would be impractical
- You want the recipient to always see the latest data
- You need to revoke access later without chasing down copies
Revoke access at any time:
"Revoke Robin's access to my tax-filings database"Subscription
A subscription is a persistent sync channel. When you update the source database, changes propagate to the subscriber automatically through Hypercore replication.
"Subscribe Robin to updates on my county-data database"Subscriptions are one-directional — the subscriber receives your changes but their local modifications (if any) don't flow back to you.
When to Subscribe
- Multiple people need the same reference dataset kept up to date
- You're maintaining a shared data source (e.g., a curated dataset for a research group)
- You want automated sync without manual re-sharing
Managing Shared Databases
List Active Shares
"Show me my shared databases"Check Share Status
The Nest's Sharing panel shows all active database shares, including mode, recipient, and last sync time.
Revoking Access
All sharing modes support revocation:
- Clone — Nothing to revoke. The recipient already has a copy.
- Federated Read — Immediately stops query access.
- Subscription — Stops future updates. Data already received stays with the subscriber.
Next Steps
- Sharing Guide — General P2P sharing concepts
- Data Dashboard — Query and visualize shared databases
- Data Backends — Register databases for sharing