Trace the evidence!
Hyper for LangSmith
Look at your LangSmith traces, prompts, datasets, and experiments, and run code in LangSmith sandboxes, as you.
How Hyper works with LangSmith
What Hyper reads
Hyper does not keep a copy of your LangSmith data. It reads LangSmith each time you ask. It works only with LangSmith accounts in the US region. It can read your workspaces, tracing projects, runs and traces, threads, prompts, datasets and their examples, experiments, issues that LangSmith finds, your organization's usage and billing, and your sandboxes.
What Hyper can do
Create, start, stop, and delete sandboxes and snapshots, and run shell commands in a sandbox. Hyper cannot change projects, runs, prompts, datasets, or experiments. Hyper asks you before each action. You cannot choose to always allow these actions.
Before you connect
Hyper can read every trace in your workspaces, and traces often contain the full prompts, answers, and user data that your app sent to a model. Sandboxes cost money for each second that they run, until they are stopped. A deleted sandbox or snapshot is gone for good. Anyone whose text gets into your traces, such as the users of your app, can add text that hides instructions that try to trick an assistant, so check each change before you approve it.