SpotDB

SpotDB

Instant SQL sandboxes for your AI agents

19 followers

Visit website
Give your agents a secure, containerized SQL sandbox with MCP or API access enabled with DuckDB. No infra required, quick spin-up cycle. Push any CSV — query, analyze, and move on.
SpotDB gallery image
SpotDB gallery image
SpotDB gallery image
Launch tags:Developer Tools•Artificial Intelligence•GitHub
Launch Team / Built With
AssemblyAI
AssemblyAI
Build voice AI apps with a single API
Promoted

What do you think? …

Nico Bistolfi
Maker
📌
👋 Hey PH, excited to share what we’ve been working on. We’re rethinking data access in the age of AI. That means treating databases like files — lightweight, ephemeral, isolated, and scoped. Spin them up for a task, give only the data required, and tear them down when it’s done. Why? Because AI changes the rules: - You can’t just paste CSVs into Claude — context windows choke. - You can’t just hand an LLM your production keys — that’s a breach waiting to happen. - Agents need scalable, least-privilege data access, but today setting up a database for them is still painful. That’s where SpotDB comes in. It’s a secure SQL environment on demand: ➡️ Upload a CSV ➡️ Get a containerized DuckDB with MCP or API access ➡️ Query, analyze, move on — no infra, no waiting. SpotDB is a small piece of a bigger vision: building scalable data control for the AI era, where agents and apps get exactly the data they need — no more, no less. We’d love your thoughts and feedback as we build this in public 🚀
Nadia Eldeib

@nicolasbistolfi congrats on the launch! This looks really interesting.

Chang

How does it compare with DuckDB MCP?

Nico Bistolfi
Maker

@belsandre Thank you, that's a very valid question!

DuckDB MCP exposes raw SQL directly, which creates risks if an LLM issues destructive commands or stumbles over schema setup. SpotDB takes the opposite approach with read-only, guardrailed queries, making it safe to let AI assistants or automations explore CSVs and run analytics without touching production systems.

Unlike DuckDB MCP, which requires you to prep a database file and handle schema details, SpotDB is frictionless. You can upload a CSV and start querying instantly, with no setup overhead. That makes it ideal for quick data exploration, debugging, demos, or any workflow where speed and simplicity matter more than database plumbing.

SpotDB also extends beyond MCP with HTTP and socket APIs, so it plugs into scripts, dashboards, CI jobs, and no-code tools. It’s ephemeral by default, with the option to persist snapshots to S3 when needed, and it comes with built-in observability — rate limiting, injection detection, and performance stats.

This combination makes it well-suited for collaborative, production-adjacent, or enterprise scenarios where security and guardrails are non-negotiable.

Chang

@nicolasbistolfi thank you and congrats on the launch! I think I can use DuckDB MCP with read-only. Also reading a csv into DuckDB is only one line so it’s pretty quick. How is SpotDB different then? I’m curious about the “plugs into scripts and dashboards” comment and would love to understand how it would work or see some examples.

Andreea Georgescu
Maker

@nicolasbistolfi  @belsandre thanks for digging into this! You can get a lot done with read only access.

Here's an example of what you can do with SpotDB - it's read only on the original csv, but you can create new datasets in the same DB to augment and get more out of the analysis. This combined with the snapshot ability helps you share/pick up the data or trace back changes, which we've seen can help with workflows.

Chang

@nicolasbistolfi  @andreea_georgescu1 would love to try it out when your brew tap comes online! :)