datasette-agent 0.3a0

2026-06-15 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, quick

Summary

Datasette-agent 0.3a0, released on June 15th, 2026, introduces significant enhancements for its LLM-powered agent for Datasette. A new "execute_write_sql" tool now enables the agent to request user approval before writing to a database, respecting user permissions. This allows for operations like inserting data, exemplified by adding pelican sightings to a "pelican_sightings" table. Furthermore, the "datasette agent chat" terminal mode has been updated to support these user approvals. It also gains new command-line options, including "--root" for root execution, "--yes" to auto-approve questions, and "--unsafe" which combines both. This enables direct database modification through natural language prompts, such as "create a notes table" or "add a note about X", using models like gpt-5.5.

Key takeaway

For Data Scientists or AI Engineers building LLM-powered data tools, datasette-agent 0.3a0 offers a controlled method for agent-driven database modifications. You can now implement conversational interfaces that directly update databases, ensuring data integrity through explicit user approval. Consider using the "--unsafe" option for development or trusted environments to streamline automated data operations, but always prioritize permission checks in production. This release simplifies creating interactive data management solutions.

Key insights

datasette-agent 0.3a0 integrates LLM-driven database write operations with explicit user approval and permission controls.

Principles

• LLM agents can perform database writes.
• User approval is critical for data modification.
• Permissions should gate agent actions.

Method

The "execute_write_sql" tool generates SQL statements, prompts for user confirmation, and executes approved writes, considering database permissions. The "datasette agent chat" CLI supports this interaction.

In practice

• Use "execute_write_sql" for agent-driven data entry.
• Employ "--unsafe" mode for automated database changes.
• Integrate gpt-5.5 for direct database chat.

Topics

• Datasette
• LLM Agents
• Database Management
• SQL Write Operations
• User Approval
• CLI Tools
• GPT-5.5

Code references

• datasette/datasette-agent

Best for: AI Engineer, Machine Learning Engineer, Data Scientist

Related on AIssential

• datasette-agent 0.1a3 — An LLM-powered agent for Datasette improves user experience and robustness through UI and response handling enhancements.
• datasette-agent 0.1a4 — LLM agents can integrate deeply into existing data tools for conversational data exploration.
• Stop AI Agents From SQL Injecting Your Database — Secure database tools require constraining agent access and pre-approving SQL to prevent data breaches and ensure reliability.
• The Newsroom Protocol: Why Autonomous AI Agents Are a Ticking Time Bomb in B2B Databases — Autonomous AI agents with direct database write-access are a critical security and operational risk due to their probabilistic nature and susceptibility to prompt injection.
• Data Flow Control: Data Safety Policies for AI Agents — Data Flow Control (DFC) enforces tuple-level data safety policies directly within DBMS queries, ensuring compliance beyond mere correctness.
• the history of agents.md, the problems with agents.md and what makes a good one — Effective LLM agent configuration requires model-specific, concise "agents.md" files to optimize behavior and context window usage.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.