Datasette Agent, a new conversational AI assistant for the Datasette database platform, enables developers to ask questions about their data using natural language, with the ability to run entirely on local models through tools like LM Studio. The tool represents a significant convergence of two open-source ecosystems: LLM (a Python library for working with language models) and Datasette (a platform for exploring and publishing data). Released in May 2026, Datasette Agent demonstrates how developers can build AI-powered data interfaces without relying on cloud APIs or external services.

What Makes Datasette Agent Different From Cloud-Based Query Tools?

Datasette Agent provides a conversational interface that translates natural language questions into SQL queries executed against SQLite databases. In a demonstration, when asked "when did Simon most recently see a pelican?", the agent generated the appropriate SQL query and returned the correct answer: a California Brown Pelican sighting recorded on May 20, 2026. The key difference from cloud-based alternatives is that developers can run the entire system locally, keeping data private and avoiding per-query API costs.

The system works with multiple model options. The live demo at agent.datasette.io runs on Google's Gemini 3.1 Flash-Lite model, which Willison notes is "cheap, fast and has no trouble writing SQLite queries." However, developers can also run Datasette Agent against open-weight models directly on their machines. Willison demonstrated running the plugin against Gemma 4 26B, a 26-billion-parameter open-source model, using a single command in LM Studio on a Mac.

How Does the Plugin Architecture Extend Datasette Agent's Capabilities?

One of Datasette Agent's most powerful features is its extensibility through plugins. The system already includes several working plugins that expand what the agent can do beyond basic SQL queries:

• Datasette-agent-charts: Adds data visualization capabilities powered by Observable Plot, allowing the agent to generate charts directly from query results
• Datasette-agent-openai-imagegen: Integrates image generation using ChatGPT Images 2.0, enabling the agent to create visual content alongside data analysis
• Datasette-agent-sprites: Provides tools for executing code in Fly Sprites persistent sandboxes, extending the agent's computational reach

Willison noted that building plugins is straightforward enough that AI coding assistants like Claude Code and OpenAI Codex can write them with minimal guidance. This extensibility means developers can customize Datasette Agent for domain-specific tasks, from financial analysis to scientific data exploration.

What Technical Requirements Does Datasette Agent Need to Function?

Datasette Agent requires models capable of reliable tool calling and SQL query generation. Willison explained that "Datasette Agent needs reliable tool calls and the ability for a model to produce SQL queries that run against SQLite." The good news for developers is that open-weight models released in the past six months have increasingly demonstrated these capabilities, making local deployment more practical.

Willison

This technical requirement shaped the design of the system. Models must not only understand natural language but also generate syntactically correct SQL and understand when to invoke specific tools. The improvement in open-source model quality means developers are no longer forced to choose between capability and privacy; they can now have both.

Steps to Get Started With Datasette Agent Locally

• Install LM Studio: Download and set up LM Studio on your machine to run open-source language models locally without cloud dependencies
• Load a Compatible Model: Select and download a model like Gemma 4 26B that demonstrates strong SQL generation and tool-calling abilities
• Configure Datasette Agent: Use the provided command-line interface to point Datasette Agent at your local model and specify your SQLite database
• Explore Your Data: Start asking natural language questions about your data and iterate on the results to refine your queries
• Extend With Plugins: Add plugins like datasette-agent-charts to unlock additional capabilities tailored to your use case

What Does This Mean for the Broader AI Ecosystem?

Datasette Agent's release signals a maturation of the local AI development ecosystem. Willison noted that the project "opens up so many opportunities for the LLM and Datasette ecosystem in general." The convergence has already influenced development of the underlying LLM library, with a major 0.32a0 refactor underway that extracts agent abstractions from Datasette Agent itself.

Willison

The project also demonstrates a practical use case for on-device AI that goes beyond chatbots or code generation. Data exploration and analysis represent a genuine business need, and Datasette Agent shows how developers can build production-grade tools that keep sensitive data local while leveraging AI capabilities. Willison plans to use Datasette Agent to build "Claw," a personal AI assistant that aggregates data from different parts of his digital life, suggesting the tool's potential for both individual developers and enterprises managing proprietary information.

Datasette Cloud users will also gain access to Datasette Agent, expanding its reach beyond developers running local instances. The project includes a dedicated Discord channel for community discussion and collaboration, indicating that Willison expects significant developer interest in building on top of this foundation.