TLDR;

Inconvo helps companies add chat-with-data agents to their product. We provide a platform for configuring agent rules and directions and a simple API endpoint to call them. We’re open source — check out our repository here. We also offer Inconvo cloud — check it out at app.inconvo.ai

https://www.youtube.com/watch?v=sbDTAbVG-WQ

The problem:

Today, many product teams are under pressure to add “data agents” to their apps that connect to live databases and answer user questions.

Chat-with-data demos are easy. Shipping one to real customers is hard.

Most teams start with “LLM generates SQL + prompts,” then run into three issues:

• Data safety: you can’t reliably stop the model from pulling the wrong rows, crossing tenants, or selecting sensitive columns.
• Business correctness: metric definitions, allowed filters/enums, join paths, and formats get split across prompts which can confuse the model and drift over time.
• Boundaries: because the model can generate arbitrary SQL, “only allow these joins/filters/columns” becomes brittle after-the-fact validation.

The result is a second system of guardrails and checks just to make it safe to ship.

The solution:

Inconvo flips the approach: constrain the agent up front instead of trying to rein it in after.

• A semantic model as the contract: define approved tables/columns, metrics, filters/enums, and join paths in one place.
• Constrained query plans (not SQL): the model outputs structured query plans which are validated against your semantic model before execution.
• Automatic access rules: tenant scoping and field restrictions are enforced on every request with deterministic code not prompts.

You can call your data agent from your application in just a few lines of code:

Our ask:

Check out our repo and leave us a ⭐ on Github: https://github.com/inconvoai/inconvo

Run Inconvo OSS locally npx inconvo@latest dev Or sign up to our cloud for free https://app.inconvo.ai