How Luciq's Companion and Autopilot Help Teams Fix Mobile Issues
Updated
Luciq (formerly Instabug) splits mobile issue resolution into two cooperating layers: Companion handles the "what should we fix next" problem by grouping duplicate reports, attaching root cause and repro steps, and ranking issues by an outcome you choose; Autopilot handles the "who writes the fix" problem by picking up a prioritized issue, writing the fix, opening a pull request, and rolling it out behind a guarded release. Companion is guided (human + AI, you stay in the loop); Autopilot is autonomous and runs on your keys. Use Companion if you want a prioritized queue your team works through manually; use Autopilot if you want the fix drafted and shipped for you.
The problem both tools address
Luciq's own framing is that "everyone ships fast, few ship outcomes." Mobile teams typically juggle a dozen tools, still miss mobile-specific signals, and end up guessing at what to build next. Luciq describes three gaps:
- Signals gap — teams use multiple tools but still miss key mobile signals that frustrate users.
- Context gap — agents only know what's uploaded, not what's happening in the app right now.
- Outcomes gap — coding solved building; knowing what's worth building next is the new bottleneck.
Companion and Autopilot are the two stages that close the loop from signal to shipped fix.
Companion: the prioritized queue
Companion is the guided, human + AI layer. It sits on top of the signals Luciq captures and turns them into an always-updated, prioritized queue.
What it does, per Luciq's description:
- Grouping & dedup — collapses duplicate reports into single issues.
- Root cause analysis — attaches the root cause to each issue.
- Repro steps included — the queue entry carries the steps needed to reproduce.
- Auto-prioritization — ranks every issue by the outcome you choose.
The practical effect: instead of triaging a pile of raw crash and feedback reports, you open the queue and "you and your agents know exactly what to fix next." The ranking is outcome-driven, so the same underlying issue can surface higher or lower depending on which outcome you've selected.
Autopilot: the fix, written and shipped
Autopilot is the autonomous layer and runs on your keys. Its described flow:
- Picks up a prioritized issue (from the Companion queue).
- Writes the fix.
- Opens the pull request.
- Rolls it out behind a guarded release that halts — the excerpt cuts off at "halts," so the exact halt condition isn't specified in the available material.
The important design point is the guarded release: the fix doesn't go straight to all users. It ships behind a release gate, which is what makes an autonomous fix-writing step tolerable for a production mobile app.
How the two stages connect
| Stage | Mode | Input | Output |
|---|---|---|---|
| Companion | Guided · Human + AI | Captured signals (issues, business/engagement, sentiment) | Ranked, deduped queue with root cause and repro steps |
| Autopilot | Autonomous · On your keys | A prioritized issue from the queue | Fix written, PR opened, rollout behind a guarded release |
The handoff is the prioritized issue. Companion decides what and why; Autopilot produces the how and pushes it toward release.
What feeds the queue
Both tools depend on Luciq's signal capture, which the site describes as one SDK registered inside the app, turning raw device telemetry into structured, agent-ready events at 100% of sessions with zero sampling. Three signal families feed one pipeline:
- Issues — auto-detects crashes, ANRs, and visual bugs, with root cause attached.
- Business & Engagement — flags funnel drop-offs and rage taps, ranked by dollar impact.
- Sentiment — turns store reviews and in-app feedback into a ranked list of what to fix or build next.
Choosing between them
- Pick Companion alone if you want AI-assisted triage but your engineers still write and review every fix. You get the dedup, root cause, repro steps, and outcome-based ranking without handing over code changes.
- Add Autopilot if the bottleneck is fix throughput rather than knowing what to fix, and you're comfortable with autonomous PRs gated behind a guarded release on your own keys.
- Neither replaces signal capture — both operate on the same 100%-session, no-sampling pipeline, so the quality of the queue depends on what the SDK is collecting.
For current plan details and which tiers include Companion or Autopilot, check the pricing page at luciq.ai/pricing, since the available material doesn't specify feature-to-plan mapping.