50 plan calls / day
The
plan tool. Direct SQL execution (run) is counted separately.500,000 tokens / day
LLM token budget for planning. Consumption scales with query complexity and schema size.
500 tables · 3,000 columns
Hard limit per database connection. Scope the connection to a specific schema to stay under the limit.
300 rows per query
Rows returned to your AI client. The Snowflake query runs in full; DataFrey caps what it sends back.
Planning & indexing
Planning requires a built index — rundatafrey index after connecting your database.
| Constraint | Value |
|---|---|
| Max tables indexed per connection | 500 |
| Max columns indexed (across all tables) | 3,000 |
| Daily plan requests | 50 |
| Daily reindex requests | 3 |
| Daily LLM token budget | 500,000 tokens |
| Per-plan token limit | 100,000 tokens |
| Per-table index timeout | 60 s |
| Total reindex timeout | 14 min |
Query execution
| Constraint | Value |
|---|---|
| Max rows returned | 300 |
| Statement timeout | 30 s |
| Max query length | 10,000 characters |
| Daily run requests | 500 |
Database connection
Snowflake is the only supported database today — more are coming. For privileges, setup SQL, and auth, see Snowflake setup.MCP server & AI clients
Claude Code
Run
datafrey client claude to install the DataFrey plugin (MCP server + /db skill).Cursor
Run
datafrey client cursor to register MCP via deeplink.Any MCP-compatible client
Run
datafrey client mcp to print the JSON config, then paste it into your client’s MCP settings.DataFrey login session
These limits apply to your DataFrey login session, not Snowflake credentials.| Constraint | Value |
|---|---|
| Device code window | 15 min to complete browser login |
datafrey db connect freshness | Session must be ≤ 5 minutes old |
CLI
- Python 3.13 or later — required for all packages
- Install with
pip install datafreyoruv tool install datafrey
macOS
macOS Keychain
Linux
GNOME Keyring or KWallet — must be installed and unlocked before you run the CLI.
Windows
Windows Credential Locker