Promptic
Trace, evaluate, and optimize prompts and AI agent configurations
Turn agent conversation traces into tested improvement pull requests

Basalt analyzes conversations from deployed AI assistants and proposes improvements to their behavior. It connects to observability sources such as Langfuse, Braintrust, or Datadog, then groups traces to identify recurring failures. Those patterns inform changes to system prompts, tool definitions, or examples.
Suggested fixes are checked against a team's golden dataset before a pull request is opened, giving developers a concrete change to review. The workflow is designed to repeat overnight so recent user interactions feed the next iteration. Basalt is useful for teams operating AI assistants that want production conversations to become actionable engineering work, with humans deciding which changes to merge.
Tried it? Share a useful thought or ask a question.
Good conversations start with a hello.
Sign in to comment