We hosted a live launch preview, “Level Up Your Agentic Workflow,” to introduce two things at once: GitKraken Insights for Developers and Kepler. Jeremy Castile, our VP of Developer Research, opened with research on how teams are actually using AI agents today. If you want the full breakdown of that data, including the adoption numbers and the gap between feeling faster and proving it, we published it separately in State of AI in Engineering 2026: The Proof Gap. This post covers what came after: the live product demos and the questions that came out of them.

GitKraken Insights for Developers

Jeff Schinella, product director on GitKraken Insights, walked through the platform live. Insights already gives organizations visibility into AI adoption, effectiveness, and cost. What’s new is putting that same visibility directly in developers’ hands, not just engineering leadership.

Inside Insights for Developers, you get:

• Adoption and cost breakdowns by tool, including token consumption estimates, model usage (Opus 4.8, Opus 4.6, and more), and request counts, benchmarked against your org’s 90th percentile and average
• Reliability and health data, including cache hit rates (switching models mid-session costs you reprocessed context), tool call failures, and rate limit hits
• Repository readiness scoring, checking for AI instructions files (CLAUDE.md, Copilot instructions), CI/CD rules, a strong README, branch protection, linter and test config, and shared skills your team can reuse
• Contribution and PR-level mapping, showing exactly which agent sessions, models, and skills went into a given pull request or commit, and what each one cost
• AI-driven coaching, recommending which model fits which task based on cost, efficiency, and error rate from your own sessions

Jeff walked through a real pull request during the demo, one that cost over $100 relative to others in the same repo, and showed how you can drill into which skills ran, which models were used, and where the cost actually came from. That level of detail, tied to an individual PR, is the part that’s new here.

Support for Claude Code is strongest today, with Codex and other tools coming. Insights for Developers is live now in preview.

Kepler

Justin Roberts, Senior Director of Product, demoed the second product. Teams weren’t struggling to start using agents. They were struggling to build consistent workflows once agents were scattered across CLIs, IDE sessions, and desktop apps, each developer working single-threaded.

Kepler is agent and model agnostic, connecting to Claude, Codex, Copilot, and Cursor. At its core is the task: point Kepler at a repository, pick an agent and model, send a prompt, and Kepler handles the Git work, branches and worktrees included, so developers don’t have to think about it.

The real value shows up when Kepler connects to issue trackers like Linear, Jira, and Azure DevOps. Pull in assigned issues, pick one or several, and launch tasks in parallel, each in its own worktree on the same repository. Kepler also surfaces terminals for running commands and shows the code an agent is producing in real time, and it supports pull request review, pulling PR context and comments into a task so an agent can address feedback and push changes back.

From the Q&A

A few things attendees asked about, and what Justin said:

• Can tasks talk to each other? Not yet, though it’s a direction being considered. Today, a single task can span multiple repositories and issues at once, which covers a lot of the same ground.
• What about dev containers? Kepler works today with remote environments over SSH or WSL. Kepler-hosted containers for teams without that setup already in place are under consideration but not built yet.
• Will there be an editor for manual intervention? Likely some basic file editing is coming, so developers can make quick edits to what an agent produces. The team is intentionally stopping short of building a full IDE.
• What about worktree disk usage building up over time? Cleanup is manual today, you can delete branches, tasks, and worktrees yourself. More automated cleanup tied to merge and PR-closed events is in progress.
• Any plans for automations? Yes, the next step after the current actions system is letting teams chain actions into an end-to-end workflow, research, plan, implement, self-review, PR review, merge, customized to how each team actually works.

Kepler connects to Claude Code today via OpenTelemetry hooks and events, which is also how GitKraken Insights pulls in Claude Code usage data.

Try it

Engineering leaders can get GitKraken Insights for their whole team, which includes Insights for Developers for every seat, at gitkraken.com/insights. Any developer can download Kepler for free at gitkraken.com/kepler and start running parallel agents today.