The bugs your users hit, fixed before they report them.
Every morning, Kavela goes through your session replays, finds what broke and opens the fix as a pull request.
99.8% of session recordings are never opened by anyone. Kavela goes through yesterday's every morning.
PostHog, Nobody watches session replays (2026)Running in under two hours. No integration to write.
Describe the job in plain words. Kav finds the apps it needs, asks you to connect each one once and sets up the workflow.
- One message to start
- The workflow, the app connections and the agent all come from what you said.
- One click per app
- PostHog, GitHub and Slack or Lark each arrive as an approval, shared by every agent in the workspace.
- Any tool you use
- Anything with MCP, OAuth 2.0, an API key or a sign-in page.
Know what it will do, step by step.
Your sentence is drawn as steps, each with the app it works in. Change it by saying what should change.
- On your schedule
- Every morning, when something happens in your tools, or when you press Run.
- Every change is a version
- See what changed, when, and how each version scores.
- Readable by the whole team
- Product, QA and engineering read the same steps in plain words.
A fix reaches you only after it passes your checklist.
A second agent that didn't write the fix checks it against your goal and every line of your checklist.
- Sent back, not shipped
- A fix that falls short goes back with the exact gap. After three tries, you're told what's missing.
- Measured across runs
- Each version has a pass rate, and no better or worse claim until both have ten checked runs.
- Told when it keeps missing
- A line that fails twice is raised with you, with the line it misses most.
Tune it for a week. Then it runs on its own.
Correct it in plain words. What you settle is kept as its memory and playbooks, so each run starts from the last.
- Rules you can read
- Its memory lists every rule it follows, and you can change any of them.
- Playbooks from the source
- Guides for the tools it works in, written by their makers.
- As much autonomy as you allow
- Supervised asks before anything leaves your workspace. Autonomous asks once, then runs.
Common questions.
Does it replace our end-to-end tests?
No. Tests check what you thought to test. This goes through what real users hit, and opens a pull request when it finds the cause.
Will it merge code on its own?
No. It opens a pull request and your team merges it. On Supervised it asks before it opens one; on Autonomous it asks the first time, then runs.
Which replays does it go through?
The ones you describe, like yesterday's sessions that hit an error or a rage click. PostHog's MCP server lets it search recordings by events, behaviour and time.
Do we need PostHog?
This job uses PostHog. Any tool with MCP, OAuth 2.0, an API key or a sign-in page can take its place.
Slack or Lark?
Either, or both. It posts where your team already works.
How long does it take to set up?
TrueNorth's was running in under two hours. The week after went to tuning it, and now it runs on its own.
How much does it cost?
Every plan starts with a 14-day free trial, no card. Bring your Claude or Codex subscription at no extra cost, or use the included Kav AI usage.
Start with one task.
Delegate a task free14-day free trial, no card. Bring Claude or Codex, or try included Kav AI usage.
Claude
Codex
Kav AI