Repository navigation
Tracking: FlyBot vs Grok Bot capability gaps (connectors, 24/7, mobile, team, quality) #361
Description
Why
A working comparison of FlyBot vs Grok Bot (product capability, not a code audit). Use this issue to track the gap, decide what to chase, and link child issues. Discussion welcome; nothing here is a commitment to copy Grok Bot.
Where FlyBot is ahead (do not dilute)
- Multi-Bot collaboration: tree, Bot Social, face-to-face rooms, per-Bot grants.
- Visible collaboration: Timeline swimlanes, approval cards, Agent inbox ([Feature RFC] Agent inbox board: one place for tasks, approvals and failures (personal first, team second) #349).
- One isolated instance per person (Deployment step 1: official image and one isolated instance per person #352), natural for team mode.
Large gaps (Grok Bot is more mature)
| Area | Gap | Child issues |
|---|---|---|
| Connector / App ecosystem | Catalog, one-click install, OAuth, tools appear after auth | #359 |
| Proactivity and 24/7 | Routines, runs while the user is away | #353 #354 #355 |
| Cloud residency and computer use | Hosted machine, browser desktop, cloud coding agent | #352 (first step only) |
| Team and governance | Multi-agent teammates, shared memory, draft-before-send, audited outbound | #51 #55 #56 #57 #58 #149 |
| Onboarding and proof of quality | Open-box connectors; Nightly aligned to user journeys; perf baseline | #358 #356 #357 #351 |
| External channels | Slack etc. as first-class speak/listen paths | #355 |
| Mobile app | No iOS/Android (or strong mobile web) entry; blocks anytime approvals and reminders | new: see below |
Mobile (new, important)
FlyBot is Desktop / web today. Without a phone entry point, proactive reminders (#353–#355) and on-the-go approvals stay weak. Proposed child work (open as separate issues when ready):
- Phase A: mobile-friendly web (PWA): push for approvals and reminders, Agent inbox ([Feature RFC] Agent inbox board: one place for tasks, approvals and failures (personal first, team second) #349) usable on a phone.
- Phase B: native or wrapped app (iOS / Android) with reliable push, biometric unlock for sensitive approve, deep links into cards.
- Hard dependency: a reachable personal instance (Deployment step 1: official image and one isolated instance per person #352); the phone is a client, not a second source of truth.
Three highest-leverage bets (from earlier discussion)
- Plugin and App connect: catalog, guided auth, and a fast developer scaffold (cf. Grok Bot connectors) #359 connectors / Apps — daily work plugs in quickly.
- Deployment step 1: official image and one isolated instance per person #352 one instance per person + Proactive reminders 1/3: event reminders (approval waiting, failed, finished) in app and desktop #353 event reminders — works with the laptop closed, comes to find you.
- User-scenario matrix: map journeys to Nightly / live / perf coverage #358 user-scenario Nightly + Performance dashboard: Nightly CI metrics for FlyBot smoothness #357 performance board — quality is visible every morning.
Differentiation stays multi-Bot collaboration and Timeline: win on “team of Bots works better”, not on “same as Grok Bot”.
How to use this issue
- Comment to add / drop / re-rank gaps.
- Open child issues for mobile phases and any missing rows; link them in the table.
- Do not implement from this issue directly.
Process
RFC and tracking only. Implementation of children via OpenCode + Codex review per team rule.
Activity
hsliuustc0106 commentedon Oct 18, 2026
Child opened for mobile: #362 (PWA first, then native).
hsliuustc0106 commentedon Oct 18, 2026
Mobile platform suite opened: #363 (iOS + HarmonyOS + Android). Complements #362 (PWA / native product phases).
hsliuustc0106 commentedon Oct 18, 2026
Related gap/child: #364 — per-child harness + model in a tree (mix backends for task quality).
hsliuustc0106 commentedon Oct 18, 2026
Correction: the tree multi-harness issue is #365 (not #364).
hsliuustc0106 commentedon Oct 18, 2026
Follow-up on #365: pull Magpie ideas (harness ≠ provider, local gateway, routing groups, profiles) — comment on #365.
hsliuustc0106 commentedon Oct 18, 2026
Broader Magpie-inspired RFC opened (harness × provider, gateway, groups, profiles) — search issues for
RFC: harness × provider.hsliuustc0106 commentedon Oct 18, 2026
Magpie-inspired RFC: #366.
Birfy commentedon Oct 18, 2026
Maintenance note (2026-10-18 22:20 Asia/Shanghai): one line in "Where FlyBot is ahead" conflicts with a current owner decision. Timeline swimlanes are hidden by default and not loaded anywhere in the FlyBot UI (sidebar entry included), and they get no demo (see #350). Please don't count them as a visible strength until the owner turns Timeline back on. The approval cards and the Agent inbox idea (#349) are unaffected.
hsliuustc0106 commentedon Oct 18, 2026
Ideas from Claude-related write-up (secondary source): #373.
hsliuustc0106 commentedon Oct 18, 2026
Safety RFC from the same source: #386.