HypAware
View as Markdown

See where the work goes

Find out which tickets and pull requests your agents' work actually went into.

HypAware connects a session to the commits it made, a commit to its pull request through the GitHub integration, and a pull request to its ticket through the Linear integration. So you can start from a ticket or a PR and get every session behind it: how long it took, how many times the agent went back and forth, where it got stuck, and what it used. Or start from a session and find the ticket it was for.

Ask your agent

Start from one ticket or PR:

From my HypAware history, for [ticket], show the pull requests, the sessions behind them, and where the agent got stuck.

From my HypAware history, which sessions went into [PR], how long did they take, and how many tokens did they use?

Look across a sprint or a month:

From my HypAware history, which tickets this sprint took far more agent work than the rest, and why?

From my HypAware history, which pull requests this month needed the most sessions, and what made them hard?

From my HypAware history, which pull requests needed the most rework after review?

Go the other way:

From my HypAware history, which ticket and pull request was this session for?

From a terminal:

Terminal
hyp ask "From my HypAware history, which tickets this sprint took far more agent work than the rest, and why?"

Before you start

  • Pull requests need the GitHub integration. Run hyp github connect to turn it on and sign in; hyp setup also offers it at the end. Use hyp github backfill to bring in history from before you turned it on.
  • Tickets also need the Linear integration. Run hyp linear connect to sign in and choose your Linear teams. It links an issue to the pull requests attached to it in Linear, and collects on the host where you set it up.

What you get back

For each ticket or pull request: the sessions that worked on it, when they ran, how much back-and-forth they took, token use where the client records it, and a short account of what went smoothly and what didn't. Across a sprint, the agent ranks tickets or PRs by the work that went into them and explains the outliers.

Get a better answer

  • The link runs through commits. Work that never produced a commit, such as exploration or an approach you threw away, doesn't link on its own. Ask the agent to also match sessions by branch name, which every session records.
  • A ticket links only to pull requests attached to it in Linear. If your team doesn't attach PRs to issues, start from the PR instead.
  • Token counts depend on the client. Some integrations don't record token use; see integrations.
  • An outlier is a question to ask, not a verdict. Ask why a ticket took so much work: a vague spec, a flaky test, and a missing tool look the same in a count and very different in the sessions.
Was this page useful?

↑↓ navigate↵ openesc close