Ask questions of your codebase and get a written answer

One assistant, set up once, that anyone on the team can ask. It reads the code in a sandbox and produces an answer or a file you can keep.

How it runs.

Starts on: a question in a conversation, or a scheduled report.

  • Reads the repositories

    The assistant works on the code in an isolated sandbox, with the repository access you granted it.

  • Does the analysis

    Dependency maps, dead code, migration inventories, onboarding guides: the work that is useful and never urgent.

  • Produces something you can keep

    A written answer, a document, or a spreadsheet, saved where the team can find it.

What your team controls.

The instructions are yours, and so are the limits.

  1. A record of every run

    See what the assistant read, what it did, and what it produced, while it works and after it finishes.

  2. The model you choose

    Pick a model for each assistant and change it later. A cheaper model for routine passes, a stronger one where it matters.

  3. Spend you can see

    Each run records what it cost. Usage breaks down by person, assistant, and routine, so the bill is never a surprise.

  4. A person approves what matters

    Set the assistant to ask before it writes to a connected account. It waits for the person the work runs for.

Start with one useful job. Make it work for your team.

Connect a repository, write the instructions, and run your first assistant.

Free to start. No credit card.