datasette-agent 0.3a0

Simon Willison · Simon Willison · 2026-06-15

Simon Willison releases datasette-agent 0.3a0, adding an `execute_write_sql` tool that prompts users for approval before writing to databases, along with new CLI options for unsafe auto-approval and root access in the chat terminal mode.

Open original ↗

Appears in

Simon Willison's AI-Augmented Datasette Ecosystem: Agent, Apps, and Plugins

Extraction

Topics: datasettellm-tool-useai-agentsdatabase-toolsopen-source

Claims

datasette-agent 0.3a0 introduces an `execute_write_sql` tool that requires explicit user approval before executing write operations against a database.
The new `--unsafe` flag auto-approves all prompts, enabling direct database modification via natural language in the CLI.
Tools in datasette-agent can now provide plain text output alternatives to HTML for display in the CLI.
The chat terminal mode now supports `--root`, `--yes`, and `--unsafe` flags for fine-grained control over permissions and approval behavior.

Key quotes

The new execute_write_sql tool can now prompt the user for all kinds of useful operations.

The datasette agent chat content.db -m gpt-5.5 --unsafe command can now be used to chat directly with a specific database and directly modify it through prompts like 'create a notes table', 'add a note about X' etc.