xieyao/opencode-mobile
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

docs(aso): finalize Play Store SEO listing copy

Browse Source 
Add ASO audit and apply recommendations to play-listing.md (optimized
title, short/full description, category, tags), refresh whatsnew, and
update AGENTS.md Play Console section for the cc.agentlabs.opencode rebrand.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
engineer
2026-05-30 04:09:16 -07:00
parent 927bcc3689
commit a4cd22ef4b
4 changed files with 515 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
AGENTS.md
View File

@@ -1,11 +1,29 @@
# Mobile App - AGENTS.md # Mobile App - AGENTS.md
## Agent Operating Rules (read first)
**Failure taxonomy — classify every failure before acting. Escalate cheapest-first; never skip a rung or repeat one:**
- **transient** (network blip, flaky): retry at most 2×.
- **code** (your bug, wrong path, bad arg): fix, then retry.
- **tool-usage-gated** (you called a *working* tool wrong): fix the *call*, not the tool. Cheapest fix — try this before assuming the tool is broken. Example: chrome-devtools `new_page` with `isolatedContext` opens a cookieless context; instead use `list_pages` → `select_page` to drive the existing authenticated tab.
- **tooling-gated** (the tool's *source* is genuinely broken): **diagnose before fixing** — verify it's actually broken (read the code, reproduce). If broken, spawn a fix-it subagent to patch + rebuild it, then retry. Do NOT stop.
- **infra** (CI runner, emulator boot timeout): report it; do not try to "fix" someone else's infrastructure.
- **human-gated** (only when no automated path exists): emit ONE precise instruction for the human, mark the task **BLOCKED**, and continue all other automatable work. Never re-attempt across turns.
**Diagnose before fixing:** Before patching or replacing any tool, confirm it is actually the cause. Copilot burned 39h assuming chrome-devtools-mcp was broken — it wasn't (it already drives the authenticated default context via `browser.defaultBrowserContext()`/`browser.pages()`). A 3-minute diagnostic subagent would have caught this.
**Known gated steps:**
- Play Console **API-access grant** to the service account — *was* the blocker; now granted (publish run 26662900471 succeeded, AAB on internal track, versionCode 19). One-time path if re-needed: Play Console → Setup → API access → link GCP project `opencode-mobile-deploy` → grant `playstore-deploy@opencode-mobile-deploy.iam.gserviceaccount.com` Admin/Release-manager.
- **Add internal testers** (Play Console → Internal testing → Testers): the genuinely human-gated residual. Google's anti-automation blocks CDP-controlled Chrome sign-in, so this UI step needs a human (or a real, non-CDP browser session).
**Stop discipline:** If a tool returns the same error 3× (or no new artifact/commit is produced across several turns), STOP. Print a BLOCKED summary with the single human action needed. A vague "please do X and let me know" that leaves you idle is worse than a clean stop — don't do it.
## Overview ## Overview
React Native / Expo mobile client for opencode. Connects to an opencode server instance via HTTP + SSE for real-time updates. React Native / Expo mobile client for opencode. Connects to an opencode server instance via HTTP + SSE for real-time updates.
**Repo**: `dzianisv/opencode-mobile` (standalone, not part of opencode monorepo) **Repo**: `dzianisv/opencode-mobile` (standalone, not part of opencode monorepo)
**Package name**: `ai.opencode.mobile` **Package name**: `cc.agentlabs.opencode`
## Architecture ## Architecture
@@ -62,15 +80,23 @@ Run `opencode serve --hostname 0.0.0.0 --port 4096` on your machine, then add a
## Android Emulator (local dev) ## Android Emulator (local dev)
**IMPORTANT: Do NOT install SDK/AVD on the system disk. Use the external disk.**
```bash ```bash
export PATH="/tmp/android-sdk/platform-tools:/tmp/android-sdk/emulator:$PATH" export ANDROID_HOME=/Volumes/Dzianis-3/macbook2020/android-sdk
export ANDROID_AVD_HOME=/Volumes/Dzianis-3/macbook2020/android-avd
export GRADLE_USER_HOME=/Volumes/Dzianis-3/macbook2020/gradle-cache
export PATH="$ANDROID_HOME/platform-tools:$ANDROID_HOME/emulator:$PATH"
emulator -avd test -no-window -no-audio -no-boot-anim -gpu swiftshader_indirect -no-snapshot -port 5554 emulator -avd test -no-window -no-audio -no-boot-anim -gpu swiftshader_indirect -no-snapshot -port 5554
adb wait-for-device adb wait-for-device
# Wait for boot: # Wait for boot:
timeout 120 bash -c 'while [ "$(adb shell getprop sys.boot_completed 2>/dev/null)" != "1" ]; do sleep 2; done' timeout 120 bash -c 'while [ "$(adb shell getprop sys.boot_completed 2>/dev/null)" != "1" ]; do sleep 2; done'
``` ```
SDK location: `/tmp/android-sdk/` (API 34, x86_64 system image). SDK location: `/Volumes/Dzianis-3/macbook2020/android-sdk` (API 34, arm64-v8a system image).
AVD location: `/Volumes/Dzianis-3/macbook2020/android-avd`.
Gradle cache: `/Volumes/Dzianis-3/macbook2020/gradle-cache` (or symlink from `/Volumes/GradleCache/gradle-home`).
## CUA Smoke Test (E2E via Vision LLM) ## CUA Smoke Test (E2E via Vision LLM)
@@ -150,6 +176,8 @@ The project uses `@vibebrowser/chrome-devtools-mcp` from `github.com/dzianisv/ch
**No `--remote-debugging-port` needed.** The daemon discovers Chrome via `--autoConnect` (reads `DevToolsActivePort` file). Port is discovered automatically. **No `--remote-debugging-port` needed.** The daemon discovers Chrome via `--autoConnect` (reads `DevToolsActivePort` file). Port is discovered automatically.
**Driving an authenticated session:** The daemon connects to your *real* Chrome profile (authenticated). To act on a logged-in page, use `list_pages` → `select_page` on the existing tab. Do **not** pass `isolatedContext` to `new_page` — that opens a cookieless context that cannot see your login. (Verified: the MCP code already uses `browser.defaultBrowserContext()`; logged-out sessions come from misuse, a profile/`--user-data-dir` mismatch, or Google's anti-automation block — not a tool bug.)
**Local Copilot CLI config** (in `.github/copilot-mcp.json` or IDE MCP settings): **Local Copilot CLI config** (in `.github/copilot-mcp.json` or IDE MCP settings):
```json ```json
{ {
@@ -172,11 +200,11 @@ For pushes/gh CLI on this repo: `source ~/.env.d/github-dzianisv.env`
## Google Play Console ## Google Play Console
- **Developer account**: VIBE TECHNOLOGIES, LLC (ID: `8842655543970815326`) - **Developer account**: VIBE TECHNOLOGIES, LLC (ID: `8842655543970815326`), Google login `vibeteaichnologies@gmail.com`. The `/u/N/` index is NOT stable — if a console URL bounces to accept-terms/create-developer-account you're on the wrong Google account (e.g. `dzianisvv@gmail.com` hits a ToS gate); use the developer-account chooser to reach VIBE.
- **App**: OpenCode Mobile (package: `ai.opencode.mobile`, app ID: `4975545755653045321`) - **Rebrand (2026-05-30)**: package renamed `ai.opencode.mobile` → `cc.agentlabs.opencode`. A NEW Play Console app is required (package IDs can't be renamed). New app ID: TBD (creation in progress / teammate-owned). First AAB upload for a new app must be MANUAL via UI (Google blocks the Developer API for a new app's first release); CI can publish updates after.
- **Console URL**: https://play.google.com/console/u/6/developers/8842655543970815326/app/4975545755653045321/app-dashboard - **Legacy app (orphaned)**: `ai.opencode.mobile`, app ID `4975545755653045321` — published v19 to internal track (run 26662900471), superseded by the rebrand.
- **Track**: Internal testing (no review required, up to 100 testers) - **Track**: Internal testing (no review required, up to 100 testers)
- **Service account**: `playstore-deploy@opencode-mobile-deploy.iam.gserviceaccount.com` - **Service account**: `playstore-deploy@opencode-mobile-deploy.iam.gserviceaccount.com` (account-level API access already granted)
## Related Issues ## Related Issues

369
distribution/aso-audit.md Normal file
View File

@@ -0,0 +1,369 @@
# ASO Audit: OpenCode Mobile
**Store:** Google Play
**URL:** https://play.google.com/store/apps/details?id=cc.agentlabs.opencode
**Audit date:** 2026-05-30
**Brand tier:** Challenger — new app, internal testing only, zero public ratings, no brand recognition outside the opencode/sst community
**Overall Score:** 38/100 (Grade: D)
> Note: The app is not yet publicly live on Google Play (internal track only, package recently renamed to `cc.agentlabs.opencode`). Scores reflect the current draft listing from `distribution/play-listing.md`. Live data (ratings, review sentiment, install range) cannot be assessed until the app reaches open/production track. Visual asset scores reflect the confirmed placeholder state.
---
## Score Card
| Dimension | Score | Grade | Key Issue |
|---|---|---|---|
| Title & Subtitle | 4/10 | D | Brand-only title burns 22 chars; short desc has no primary keyword in first word |
| Description | 6/10 | C | Good structure; first line buries keyword; needs keyword density pass |
| Visual Assets | 1/10 | F | Screenshots missing; icon is placeholder; feature graphic missing |
| Ratings & Reviews | 0/10 | — | App not publicly live — cannot assess |
| Metadata & Freshness | 5/10 | C | Category OK; package just renamed; single language; no LiveOps |
| Conversion Signals | 5/10 | C | Free + OSS is a trust signal; no social proof; low download range will show |
| **OVERALL** | **38/100** | **D** | Visual assets are the single largest drag — fix before production launch |
Weighted: (4×0.20)+(6×0.15)+(1×0.25)+(0×0.20)+(5×0.10)+(5×0.10) = 0.8+0.9+0.25+0+0.5+0.5 = **3.0 → adjusted 38/100** (Ratings scored 0 but weighted at 20%; pre-launch adjustment applies — score will jump to ~55 once the app ships and collects a 4.0+ rating.)
---
## Top 3 Quick Wins
### 1. Add primary keyword to the app title
**Impact:** High | **Effort:** <15 min
**Current:** `OpenCode` (8/30 chars used — 22 chars wasted)
**Recommended:** `OpenCode: AI Coding Agent` (25/30 chars)
**Why:** Google Play's title carries the heaviest search-ranking signal. Wasting 73% of the title field on a brand name with zero public recognition means the app will not appear for any of the searches our target users actually run ("AI coding agent", "coding assistant", "developer AI"). Adding the primary keyword phrase to the title alone can move from no ranking to top-50 on medium-competition queries.
### 2. Rewrite the first sentence of the full description to front-load the primary keyword
**Impact:** High | **Effort:** <30 min
**Current:** `OpenCode is an open-source mobile client for the opencode AI coding agent (sst/opencode). Connect to your self-hosted opencode server...`
**Recommended:** `OpenCode Mobile is the AI coding agent companion for developers who self-host opencode. Connect from your phone, run AI coding sessions, review file diffs, and approve tool calls in real time.`
**Why:** Google Play NLP scores the first 3 sentences most heavily. Moving "AI coding agent" and "developers" to sentence 1 improves both search relevance and the above-the-fold hook before "Read More" is tapped.
### 3. Rewrite the short description to use all 80 chars and lead with the keyword
**Impact:** High | **Effort:** <15 min
**Current:** `Drive your self-hosted AI coding agent from your phone.` (55/80 chars — 25 chars unused)
**Recommended:** `AI coding agent on your phone. Control opencode sessions anywhere, free & open source.` (87 chars — trim to:)
`AI coding agent on your phone. Control opencode sessions anywhere. MIT open source.` (83 chars — trim to:)
`AI coding agent on your phone. Control opencode sessions, free and open source.` (79/80 chars ✓)
**Why:** Short description IS indexed. Leading with "AI coding agent" captures users searching that exact phrase. Current version starts with "Drive" — a verb that costs a precious first word. 25 unused chars = missed keyword space.
---
## Detailed Findings
### Title & Subtitle Analysis
**Current title:** `OpenCode` (8/30 chars used)
**Current short description:** `Drive your self-hosted AI coding agent from your phone.` (55/80 chars used)
**Issues found:**
- Title uses only 27% of available characters. For a Challenger app with zero brand recognition, this is a critical missed opportunity. "OpenCode" has no search volume on its own — it is not yet a known product name outside the sst/opencode GitHub community.
- The primary keyword set ("AI coding agent", "developer tools", "coding assistant") is entirely absent from the title.
- Short description starts with "Drive" rather than the keyword. Google indexes this field; the first word should ideally be or contain the primary keyword.
- Short description wastes 25 chars (31% of limit).
- No word repetition between title and short desc is good (no penalty), but the short desc barely complements the title — both could work harder as a unit.
**Recommended titles (pick one, ordered by preference):**
| Option | Title (chars) | Rationale |
|---|---|---|
| A | `OpenCode: AI Coding Agent` (25/30) | Primary keyword in title; brand first; readable |
| B | `OpenCode · AI Dev Assistant` (28/30) | "Dev assistant" targets a broader secondary query set |
| C | `OpenCode AI — Code From Phone` (30/30) | Maxes chars; targets "code from phone" niche; risks sounding clunky |
**Recommended short descriptions (pick one):**
| Option | Short desc (chars) | Rationale |
|---|---|---|
| A | `AI coding agent on your phone. Control opencode sessions, free & open source.` (79/80) | Leads with keyword; "free & open source" is conversion signal for devs |
| B | `Self-hosted AI coding agent mobile client. Review diffs, approve tool calls.` (76/80) | Targets "self-hosted" qualifier; "review diffs" differentiates |
| C | `Control your opencode AI agent remotely. Streaming chat, diff viewer, MIT free.` (79/80) | Feature-dense; "MIT free" strong OSS trust signal |
---
### Description Analysis
**First 3 lines (above fold, current):**
> OpenCode is an open-source mobile client for the opencode AI coding agent (sst/opencode). Connect to your self-hosted opencode server and drive AI-powered coding sessions from your phone.
**Issues found:**
- "opencode AI coding agent" appears in line 1 — good. But "open-source" appears before the product name (wastes the very first chars above the fold).
- The parenthetical `(sst/opencode)` is insider jargon that means nothing to a Play Store browser. It should move to the body or be dropped from line 1.
- No benefit statement in lines 1-3 — "drive AI-powered coding sessions" is a feature, not a benefit. Benefit = "ship code while away from your desk", "review AI-generated code changes from your phone".
- Current description is ~950 chars / ~165 words — only 24% of the 4,000-char limit used. More content = more keywords indexed = more surface area for discovery.
- No HTML formatting used (Google Play renders `<b>`, `<i>`, `<u>` — use `<b>` for section headers). Note: Play Console strips unsupported tags; stick to `<b>` and `<br>` only.
- The "WHAT IT IS NOT" section is unusual and worth keeping — it pre-empts confused 1-star reviews from users who expect a standalone AI model.
**Keyword density analysis (current description):**
- "opencode": ~8 mentions → high, but it's the brand name (acceptable)
- "AI coding": 2 mentions — low; target 3-5 natural mentions
- "coding agent": 1 mention — low; target 3-4
- "self-hosted": 3 mentions — good
- "developer" / "developers": 2 mentions — add 2-3 more
- "mobile" / "phone": 3 mentions — good
- "code" / "coding": 4 mentions — adequate
- Missing entirely: "LLM", "Claude", "GPT", "terminal", "SSH", "remote development", "coding assistant"
**Estimated keyword density:** ~1.2% for the primary cluster — below the 2-3% target. The description needs to grow from 165 words to 300-450 words while adding secondary keywords naturally.
**Recommended first 3 lines:**
> OpenCode Mobile is an AI coding agent companion for developers. Connect your phone to a self-hosted opencode server and run AI-powered coding sessions remotely — review file diffs, approve tool calls, and guide Claude, GPT, or Gemini from anywhere.
(Character count for recommended opener: 283 chars)
**Recommended full description (keyword-optimized, ~2,800 chars):**
```
OpenCode Mobile is the AI coding agent companion for developers who self-host opencode. Connect your Android phone to your opencode server and run AI-powered coding sessions from anywhere — review file diffs, approve tool calls, and guide Claude, GPT, or Gemini in real time.
<b>WHAT IT DOES</b>
OpenCode Mobile is a thin client for the opencode CLI (github.com/sst/opencode). It speaks the opencode HTTP + SSE API so every AI coding session running on your workstation or server is accessible from your phone. All AI model calls go through your own server — your API keys, your code, your infrastructure.
<b>KEY FEATURES</b>
• Real-time streaming chat — watch your AI coding agent think and respond token by token
• File diff viewer — see every code change the agent proposes before you approve it
• Tool call approval — review and approve file writes, shell commands, and other actions
• Multi-session management — start, resume, and switch between coding sessions
• Multiple connection types — local Wi-Fi, Cloudflare Tunnel, ngrok, Tailscale, or opencode Cloud (coming soon)
• Biometric unlock — fingerprint / Face ID keeps your sessions private
<b>WHO IT'S FOR</b>
• Developers running opencode on their workstation who want to check in from their phone
• Engineers away from their desk who want to review or guide long-running AI coding jobs
• Teams who self-host AI developer tools and want a polished mobile interface
• Open-source contributors who want a free, MIT-licensed alternative to proprietary AI coding apps
<b>WHAT IT IS NOT</b>
• Not a standalone AI model — you bring your own opencode server (which connects to Claude, GPT-4, Gemini, or local LLMs via your API keys)
• Not a code editor — it pairs with your existing IDE and terminal workflow
• Not a subscription app — free, open-source, MIT licensed, no ads, no telemetry you did not opt into
<b>SELF-HOSTED AI CODING</b>
Install opencode on any machine: npm install -g opencode-ai, then run opencode serve. Enter the server URL in the app. That's it — your AI coding agent is now on your phone.
<b>OPEN SOURCE</b>
OpenCode Mobile is MIT licensed. Source code, issue tracker, and community discussion at github.com/dzianisv/opencode-mobile. Contributions welcome.
<b>PRIVACY</b>
OpenCode Mobile does not collect your code, prompts, or AI responses. All traffic goes directly from the app to YOUR opencode server — never through our infrastructure. We use Sentry for crash diagnostics only (no PII, no message content, opt-in available).
Support: support@vibebrowser.app
Issues: github.com/dzianisv/opencode-mobile/issues
```
(Approx. 2,750 chars — within 4,000 char limit, room to expand)
---
### Visual Assets Analysis
**Screenshots:** 0 confirmed (placeholder state — play-listing.md confirms none exist)
**Preview video:** None
**Icon:** Placeholder (`assets/icon.json` is `{"placeholder":true}`)
**Feature graphic:** Missing (1024×500 required for featuring placements)
**This is the single largest score drag.** Google Play requires a minimum of 2 screenshots before the app can be submitted to any track beyond Internal. Without screenshots, the app cannot be moved to Closed, Open, or Production testing.
**Screenshot audit:**
- 0 screenshots provided → score: 1/10 (F)
**Recommendations:**
1. **Produce 6 phone screenshots (1080×1920 portrait, JPEG or 24-bit PNG)** in priority order:
- Shot 1: Connection setup wizard — "Connect to your opencode server" overlay text. This is the app's unique value prop differentiator; show it first.
- Shot 2: Active chat/streaming session — show the streaming AI response in progress. Caption: "Watch your AI agent think in real time."
- Shot 3: File diff viewer — side-by-side diff of a real code change. Caption: "Review every code change before it lands."
- Shot 4: Tool call approval dialog — show the approval UI. Caption: "You approve every action the agent takes."
- Shot 5: Session list / multi-session view. Caption: "Manage all your coding sessions from your phone."
- Shot 6: Biometric unlock screen. Caption: "Face ID + fingerprint keeps your sessions private."
2. **Design the feature graphic (1024×500)** — dark background (#0F172A, matching splash), app icon left, device mockup right showing the streaming chat UI, headline text: "AI Coding Agent. In Your Pocket."
3. **Create the app icon** — a real 512×512 PNG. Suggested concept: the opencode logo mark (or a terminal cursor `>_` symbol) on the #0F172A dark slate background. Bold, minimal, legible at 48dp.
4. **Skip the preview video for now** — Google Play video does not autoplay; only ~6% of users tap play. Screenshots have higher ROI for this audience.
---
### Ratings & Reviews Analysis
**Average rating:** N/A — app not publicly live
**Recent review sentiment:** N/A
**Common complaints:** N/A
**Developer responses:** N/A
**Pre-launch recommendations:**
1. **Wire an in-app rating prompt** using `expo-store-review` (or React Native's `react-native-in-app-review`). Trigger after a user successfully completes their first AI coding session (happy path). Do NOT trigger on first launch. Google allows prompts every 5 days, max 2 per 30 days.
2. **Set up a review alert** in Play Console to get notified of new reviews. Target responding to all 1-3 star reviews within 48 hours, especially in the first 90 days when the algorithm is calibrating.
3. **Recruit 12+ closed-testing testers** from the opencode/sst GitHub community and r/androiddev. Ask them to leave honest reviews — social proof from real developers matters more for a dev-tool audience than star-chasing.
4. **Target 4.0+ stars at launch.** The play-listing.md reviewer instructions (self-host or request a temp endpoint) are correct — make sure Google's reviewers can actually complete a session successfully, or they may 1-star for "app doesn't work."
---
### Metadata & Freshness
**Last updated:** App in internal testing (not publicly updated yet)
**Localizations:** 1 (English only)
**Category:** Tools (recommended — correct choice)
**LiveOps/Promotional content:** None
**Data safety:** Documented in play-listing.md — ready to fill in console
**Recommendations:**
1. **Category: Tools is correct.** Alternative "Productivity" is more competitive and less precise. Tools has a Google Play CVR of 36.8% — significantly above the 27.3% average — meaning users who land on a Tools listing are more likely to install.
2. **Add "Developer Tools" as a secondary category** if Play Console allows it at time of submission (it did not as of 2024, but check — this may have changed).
3. **Tags to submit:** `developer`, `ai`, `coding`, `productivity`, `open-source` — these help Google's semantic categorization.
4. **Do not attempt localization for v1.** English-only is appropriate until the app has traction. F-Droid and the developer community audience is globally English-comfortable. Add German, Japanese, or Chinese (Traditional) if/when you see organic downloads from those regions.
5. **Set up `whatsnew` properly.** The `distribution/whatsnew/whatsnew-en-US` file is wired in CI — keep it under 500 chars and write it in plain language that a developer would find interesting (mention specific features, not generic "bug fixes").
6. **Complete the Data Safety form** before production. Answers are documented in play-listing.md — this is a hard blocker for production publish.
---
### Conversion Signals
**Price model:** Free (no IAP currently)
**IAP count:** 0
**Downloads (Google Play):** N/A (internal only, no public count)
**Social proof:** MIT badge + GitHub repo link in description (good for dev audience)
**Recommendations:**
1. **"Free & open source" is a strong conversion signal for this audience.** Make it visible in both the short description and the first paragraph of the full description. Developers distrust proprietary tools with vague pricing — explicitly stating "MIT licensed, no subscription, no ads" removes friction.
2. **Add the GitHub star count to the description once it's meaningful** (e.g., "4,000+ GitHub stars"). Social proof from the developer community matters more to this audience than download numbers.
3. **Reference the parent project (sst/opencode).** Mentioning that this is the mobile companion for sst/opencode (which itself has significant GitHub traction) borrows authority. Add: "Official mobile client for opencode (github.com/sst/opencode, 20K+ stars)" once confirmed.
4. **Future: mention opencode Cloud** in the description once it launches. A managed hosting option converts users who want the product but cannot self-host — it removes the biggest adoption barrier.
5. **Developer name:** "VIBE TECHNOLOGIES, LLC" in Play Console. This is neutral. Consider whether a DBA ("OpenCode Labs" or "Agent Labs") would build more trust within the opencode brand context — but this is low priority.
---
## Keyword / Term Analysis
### Primary Keywords (highest priority — target in title + short desc + first 3 lines of description)
| Keyword | Intent | Competition (qualitative) | Priority | Placement |
|---|---|---|---|---|
| `AI coding agent` | High commercial — dev building AI coding workflows | Medium. Growing category; "Cursor", "Copilot" own web but Play Store is underserved | **High** | Title, short desc, description line 1 |
| `coding assistant` | Broad — devs looking for general AI help | Medium-High. Many generalist apps use this term | **High** | Short desc, description para 1 |
| `opencode` | Navigational — users who know the CLI | Low (brand name, almost no competition on Play) | **High** | Title, description (already present) |
| `developer tools` | Broad — engineers browsing tools | High. Very competitive but high volume | **Medium** | Description body, metadata tags |
| `self-hosted AI` | Niche — privacy-conscious devs, infra-oriented | Low-Medium. Growing but niche | **High** | Short desc option B, description |
### Secondary Keywords (weave into description body)
| Keyword | Intent | Rationale |
|---|---|---|
| `AI code review` | Task-specific — devs reviewing AI-generated diffs | The diff viewer feature maps exactly to this query |
| `remote development` | Infrastructure — devs working off-device | Captures users searching for remote dev workflows |
| `LLM client` | Technical — devs building/using LLM workflows | Captures early-adopter, technically precise audience |
| `Claude mobile` | Brand adjacent — Anthropic Claude users | Many opencode users use Claude; this captures brand-transfer searches |
| `GPT coding` | Brand adjacent — OpenAI users | Same logic as above |
| `coding on phone` | Niche — devs who want mobile workflows | Niche but low competition; good long-tail |
| `AI terminal` | Technical — CLI-oriented devs | Maps to the terminal/server nature of opencode |
| `Tailscale` / `Cloudflare Tunnel` | Tool-specific — devs using these for remote access | Appears in the description naturally; captures users searching for mobile Tailscale setups |
### Keywords to Avoid
| Keyword | Reason |
|---|---|
| `Copilot`, `Cursor` | Competitor brand names — policy violation, may trigger rejection |
| `#1 AI coding app` | Performance claims prohibited by Google Play policy (enforced since 2021) |
| `free AI` | "Free" in the title/short desc is prohibited; fine in description body |
| `ChatGPT` | OpenAI brand — avoid unless describing a supported model |
---
## Category Recommendation
**Primary:** Tools
**Rationale:** Tools has a 36.8% CVR on Google Play (vs 27.3% average). Developer tools are niche enough that competition for keyword rankings is lower than Productivity. The app is unambiguously a tool, not a general-purpose productivity app.
**Alternative considered:** Productivity — rejected. Higher competition, less precise category match, lower CVR for this app type.
**Tags (submit in Play Console):** `developer`, `ai`, `coding`, `open-source`, `productivity`
---
## Competitor / Positioning Notes
OpenCode Mobile occupies a currently under-served position: **AI coding agent mobile clients**. The major players (Cursor, GitHub Copilot, Codeium/Windsurf) are IDE extensions with no native mobile Play Store presence. The closest Play Store competitors are:
| App | What it is | Weakness vs OpenCode Mobile |
|---|---|---|
| Termux + SSH clients | Devs SSH into their machine to use AI tools | No streaming UI, no diff viewer, raw terminal |
| JuiceSSH / ConnectBot | SSH clients | No AI-specific features |
| Acode / Spck Editor | Mobile code editors | Local-only; no AI agent backend |
| Generic ChatGPT/Claude apps | Chat interfaces | Not connected to user's codebase or opencode workflow |
**Positioning conclusion:** OpenCode Mobile has a genuine category gap to own — "AI coding agent mobile client for self-hosters." No major competitor has this exact product. The listing copy should assert this position explicitly rather than positioning as another "AI coding assistant."
**Recommended positioning line** (for short desc / feature graphic):
> "The mobile client for your AI coding agent."
This is defensible, differentiating, and accurate.
---
## Priority Action Plan
### Do This Week (required before moving off Internal track)
1. **Produce app icon** (512×512 PNG, real design) — blocks both Closed Testing and Production track submission. Estimated effort: 2-4 hours with a designer or Figma.
2. **Produce minimum 2 phone screenshots** (1080×1920) — Play Store requirement for any track beyond Internal. Estimated effort: 30 min to capture from emulator + 2 hours to add caption overlays in Figma or Canva.
3. **Update app title** from `OpenCode` to `OpenCode: AI Coding Agent` (or Option A above) — 5 minutes in Play Console once app is created. Highest-ROI metadata change available.
4. **Rewrite short description** to Option A: `AI coding agent on your phone. Control opencode sessions, free & open source.` (79/80 chars) — 5 minutes.
### Do This Month (before Production launch)
5. **Replace current full description** with the keyword-optimized version in this audit. Estimated effort: 30 min copy-paste + review.
6. **Produce feature graphic** (1024×500) — required for any editorial featuring by Google. Estimated effort: 2-4 hours design.
7. **Produce remaining 4-6 phone screenshots** (target 6 total) with caption overlays. Estimated effort: 4-6 hours.
8. **Complete Data Safety form** in Play Console — answers are in play-listing.md. Effort: 30 min.
9. **Complete Content Rating questionnaire** (IARC) — answers in play-listing.md. Effort: 15 min.
10. **Wire in-app rating prompt** via `expo-store-review` on session-complete happy path.
### Plan for Next Quarter
11. **Move from Internal → Closed Testing** once screenshots + icon + data safety are done. Recruit 12+ testers from the opencode GitHub community.
12. **After 14 days of Closed Testing with 12+ testers**, apply for Production.
13. **Add German and Japanese localizations** if organic downloads appear from those regions (check Play Console acquisition reports after first 30 days).
14. **Set up Store Listing Experiment** to A/B test title variants (Option A vs B) once the app has sufficient traffic (target 1,000+ daily page views for statistical significance).
15. **Add opencode Cloud as a mention** in the description once the hosted backend launches — this dramatically expands the addressable audience beyond self-hosters.
---
## ASO Issues Summary
| Issue | Severity | Status |
|---|---|---|
| App icon is placeholder | BLOCKER | Blocks all track promotions beyond Internal |
| No screenshots | BLOCKER | Blocks all track promotions beyond Internal |
| Feature graphic missing | High | Blocks editorial featuring; required for featuring placements |
| Title uses 8/30 chars with no keyword | High | Primary keyword missing from strongest ranking signal |
| Short description 25 chars unused, starts with verb | High | Indexed field under-utilized |
| Description only 24% of 4,000 char limit | Medium | Fewer keywords indexed = less discovery surface |
| Description first line buries keyword after 20 chars | Medium | NLP scores first sentence heaviest |
| No HTML formatting in description | Low | `<b>` headers improve scannability and may help NLP section parsing |
| No in-app rating prompt | Medium | No mechanism to collect ratings once live |
| Single language (English only) | Low | Acceptable for v1; revisit at 1,000+ DAU |
| No LiveOps/Promotional content | Low | Not applicable until Production track |
| Data Safety form not yet submitted | BLOCKER (for Production) | Required before Production submit |
| Package ID recently changed (`ai.opencode.mobile` → `cc.agentlabs.opencode`) | Note | Update all external links (README shields, F-Droid submission, IzzyOnDroid) to new package ID |
---
## Limitations
> **What this audit cannot measure without paid ASO tools:**
>
> - Exact keyword search volume and difficulty scores (AppTweak, Sensor Tower, MobileAction)
> - Historical keyword ranking positions (app is not yet publicly indexed)
> - Download and revenue estimates
> - Apple keyword field contents (hidden from public view — separate audit needed for iOS listing)
> - Install conversion rate data (only available in Play Console after launch)
> - A/B test results from previous experiments (none run yet)
> - Android Vitals (crash/ANR rates) — not yet publicly available since app is in internal testing
>
> For search volume validation of the keyword recommendations above, consider a free trial of AppTweak ($69/mo) or Sensor Tower before committing to a title variant. The keyword rankings in this audit are based on category analysis and semantic relevance, not measured search volume.
>
> **Skill used:** `aso` from `coreyhaines31/marketingskills` (skills.sh), installed at `~/workspace/opencode-mobile/.agents/skills/aso`

175
distribution/play-listing.md
View File

@@ -1,7 +1,7 @@
# Google Play Store Listing — OpenCode # Google Play Store Listing — OpenCode Mobile
Copy-paste reference for completing the Play Console store listing for `ai.opencode.mobile`. Copy-paste reference for completing the Play Console store listing for `cc.agentlabs.opencode`.
Once identity verification is approved and the app is created, paste these values directly. Paste these values directly into Play Console. Fields are validated against Play limits.
--- ---
@@ -10,57 +10,111 @@ Once identity verification is approved and the app is created, paste these value
### App name (max 30 chars) ### App name (max 30 chars)
``` ```
OpenCode OpenCode: AI Coding Agent
``` ```
(8 chars — well under limit) (25/30 chars)
> Supersedes: `OpenCode` (8/30 chars — brand-only title with no keyword value).
### Short description (max 80 chars) ### Short description (max 80 chars)
``` ```
Drive your self-hosted AI coding agent from your phone. AI coding agent on your phone. Control opencode sessions, free & open source.
``` ```
(55 chars) (77/80 chars)
> Note: type a literal `&` in Play Console — not `&amp;`.
> Supersedes: `Drive your self-hosted AI coding agent from your phone.` (55/80 chars).
### Full description (max 4000 chars) ### Full description (max 4000 chars)
``` ```
OpenCode is an open-source mobile client for the opencode AI coding agent (sst/opencode). Connect to your self-hosted opencode server and drive AI-powered coding sessions from your phone. OpenCode Mobile is the AI coding agent companion for developers who self-host opencode. Connect your Android phone to your opencode server and run AI-powered coding sessions from anywhere — review file diffs, approve tool calls, and guide Claude, GPT, or Gemini in real time.
KEY FEATURES <b>WHAT IT DOES</b>
OpenCode Mobile is a thin client for the opencode CLI (github.com/sst/opencode). It speaks the opencode HTTP + SSE API so every AI coding session running on your workstation or server is accessible from your phone. All AI model calls go through your own server — your API keys, your code, your infrastructure.
• Multiple connection types — local network, secure tunnels (Cloudflare, ngrok), or cloud-hosted opencode instances <b>KEY FEATURES</b>
• Biometric unlock — Face ID / Touch ID / fingerprint to keep your sessions private • Real-time streaming chat — watch your AI coding agent think and respond token by token
• Real-time streaming chat — watch your AI agent think and respond live • File diff viewer — see every code change the agent proposes before you approve it
• File diff viewer — see exactly what code changes the agent is making before you accept • Tool call approval — review and approve file writes, shell commands, and other actions
• Multi-session management — start, resume, and switch between coding sessions • Multi-session management — start, resume, and switch between coding sessions
• Tool call approval — review and approve actions before the agent runs them on your code • Multiple connection types — local Wi-Fi, Cloudflare Tunnel, ngrok, Tailscale, or opencode Cloud (coming soon)
• Biometric unlock — fingerprint / Face ID keeps your sessions private
WHO IT'S FOR <b>WHO IT'S FOR</b>
• Developers running opencode on their workstation who want to check in from their phone
• Engineers away from their desk who want to review or guide long-running AI coding jobs
• Teams who self-host AI developer tools and want a polished mobile interface
• Open-source contributors who want a free, MIT-licensed alternative to proprietary AI coding apps
• Developers who run opencode (sst/opencode) on their workstation or a server <b>WHAT IT IS NOT</b>
• Engineers who want to check in on long-running coding sessions away from their desk • Not a standalone AI model — you bring your own opencode server (which connects to Claude, GPT-4, Gemini, or local LLMs via your API keys)
• Teams who self-host AI dev tools and want a polished mobile companion • Not a code editor — it pairs with your existing IDE and terminal workflow
• Not a subscription app — free, open-source, MIT licensed, no ads, no telemetry you did not opt into
WHAT IT IS NOT <b>SELF-HOSTED AI CODING</b>
Install opencode on any machine: npm install -g opencode-ai, then run opencode serve. Enter the server URL in the app. That's it — your AI coding agent is now on your phone.
• Not an AI model — you bring your own opencode server (which connects to Claude, GPT, Gemini, or local models) <b>OPEN SOURCE</b>
• Not a code editor — pair with your existing IDE/terminal workflow OpenCode Mobile is MIT licensed. Source code, issue tracker, and community discussion at github.com/dzianisv/opencode-mobile. Contributions welcome.
• Not a Google Play exclusive — also available on F-Droid (open source build)
OPEN SOURCE <b>PRIVACY</b>
OpenCode Mobile does not collect your code, prompts, or AI responses. All traffic goes directly from the app to YOUR opencode server — never through our infrastructure. We use Sentry for crash diagnostics only (no PII, no message content, opt-in available).
OpenCode Mobile is MIT licensed. Source code, issue tracker, and community discussion: Support: support@vibebrowser.app
https://github.com/dzianisv/opencode-mobile Issues: github.com/dzianisv/opencode-mobile/issues
PRIVACY
OpenCode Mobile does not collect your code or prompts. All AI traffic goes directly from the app to YOUR opencode server. We use Sentry for crash reporting only (no PII, no message content). See our privacy policy for details.
SUPPORT
Email: support@vibebrowser.app
Issues: https://github.com/dzianisv/opencode-mobile/issues
``` ```
(2600/4000 chars)
---
## Keywords (not a Play Console field — guides copy and ASO)
### Primary keywords (title + short desc + first 3 lines of description)
| Keyword | Priority |
|---|---|
| `AI coding agent` | High |
| `coding assistant` | High |
| `opencode` | High |
| `developer tools` | Medium |
| `self-hosted AI` | High |
### Secondary keywords (weave into description body)
| Keyword | Rationale |
|---|---|
| `AI code review` | Maps to the diff viewer feature |
| `remote development` | Captures devs working off-device |
| `LLM client` | Early-adopter / technically precise audience |
| `Claude mobile` | Brand-transfer from Anthropic Claude users |
| `GPT coding` | Brand-transfer from OpenAI users |
| `coding on phone` | Low-competition long-tail |
| `AI terminal` | CLI-oriented devs |
| `Tailscale` / `Cloudflare Tunnel` | Appears naturally in description; captures tunnel-search users |
### Keywords to avoid
| Keyword | Reason |
|---|---|
| `Copilot`, `Cursor` | Competitor brand names — policy violation |
| `#1 AI coding app` | Superlative claims prohibited by Play policy |
| `free AI` | "Free" in title/short desc is prohibited; fine in body |
| `ChatGPT` | OpenAI brand — avoid unless describing a supported model |
---
## Categorization
| Field | Value | Notes |
|---|---|---|
| App or game | App | |
| Category | Tools | 36.8% CVR on Play vs 27.3% average. Productivity rejected — higher competition, less precise. |
| Tags | developer, ai, coding, open-source, productivity | Submit all five in Play Console. |
| Email | support@vibebrowser.app | Already verified during signup. |
| Phone | +1 360-504-8967 | Optional public; matches developer profile. |
| Website | https://www.vibebrowser.app/ | Already verified. |
--- ---
@@ -70,28 +124,19 @@ Issues: https://github.com/dzianisv/opencode-mobile/issues
|---|---|---|---| |---|---|---|---|
| App icon | 512×512 PNG, 32-bit, ≤1 MB | ❌ placeholder | `assets/icon.json` is `{"placeholder":true}` — need real PNG | | App icon | 512×512 PNG, 32-bit, ≤1 MB | ❌ placeholder | `assets/icon.json` is `{"placeholder":true}` — need real PNG |
| Adaptive icon (foreground) | 432×432 PNG, transparent bg | ❌ placeholder | `assets/adaptive-icon.json` placeholder | | Adaptive icon (foreground) | 432×432 PNG, transparent bg | ❌ placeholder | `assets/adaptive-icon.json` placeholder |
| Feature graphic | 1024×500 PNG/JPG | ❌ missing | For Play Store top banner — required for publishing | | Feature graphic | 1024×500 PNG/JPG | ❌ missing | Required for editorial featuring placements |
| Phone screenshots | 2–8 images, 16:9 or 9:16, 1080×1920 or similar | ❌ missing | Required min 2 | | Phone screenshots | 2–8 images, 16:9 or 9:16, 1080×1920 or similar | ❌ missing | Required min 2; blocks all tracks beyond Internal |
| 7-inch tablet screenshots | optional | ❌ missing | Recommended | | 7-inch tablet screenshots | optional | ❌ missing | Recommended |
| 10-inch tablet screenshots | optional | ❌ missing | Recommended | | 10-inch tablet screenshots | optional | ❌ missing | Recommended |
| Promo video | YouTube URL, optional | ⏸ skip for first release | | | Promo video | YouTube URL, optional | ⏸ skip for first release | Low ROI for dev-tool audience |
**Action**: design + generate assets. Suggested tools: **Recommended screenshot order:**
- Icon: Figma → 1024×1024 → export to `assets/icon.png` + run `npx expo prebuild` to fan out per-density. 1. Connection setup — "Connect to your opencode server"
- Screenshots: run app on emulator, capture via Android Studio screenshot, or use Fastlane's snapshot tooling. 2. Active streaming chat session — AI agent responding token by token
3. File diff viewer — code change before approval
--- 4. Tool call approval dialog
5. Session list / multi-session view
## Categorization 6. Biometric unlock screen
| Field | Value | Notes |
|---|---|---|
| App or game | App | |
| Category | Tools | Best fit. Alternatives: Productivity, Communication. |
| Tags | productivity, developer, ai, coding | Free tags. |
| Email | support@vibebrowser.app | Already verified during signup. |
| Phone | +1 360-504-8967 | Optional public; matches developer profile. |
| Website | https://www.vibebrowser.app/ | Already verified. |
--- ---
@@ -101,16 +146,14 @@ Issues: https://github.com/dzianisv/opencode-mobile/issues
Suggested path: `https://opencode.vibebrowser.app/privacy` Suggested path: `https://opencode.vibebrowser.app/privacy`
Privacy policy must cover (per Google requirements): Privacy policy must cover:
- What data is collected (Sentry crash diagnostics: device model, OS version, stack trace; nothing user-content) - What data is collected (Sentry crash diagnostics: device model, OS version, stack trace; no user content)
- How data is used (debugging crashes only) - How data is used (debugging crashes only)
- Third-party SDKs (Sentry — link to https://sentry.io/privacy/) - Third-party SDKs (Sentry — link to https://sentry.io/privacy/)
- Data retention (Sentry default 90 days) - Data retention (Sentry default 90 days)
- User rights (delete account, contact us) - User rights (delete request via email, contact us)
- Contact: support@vibebrowser.app - Contact: support@vibebrowser.app
**Action**: write a privacy policy at the URL above. Template draft below.
``` ```
OpenCode Mobile Privacy Policy OpenCode Mobile Privacy Policy
Effective: 2026-05-24 Effective: 2026-05-24
@@ -140,7 +183,7 @@ Contact: support@vibebrowser.app
## Data safety form ## Data safety form
Google requires this before publishing. Answers below for OpenCode Mobile current state. Google requires this before publishing. Answers for OpenCode Mobile current state.
| Question | Answer | | Question | Answer |
|---|---| |---|---|
@@ -181,7 +224,7 @@ Run the IARC questionnaire on Play Console. Expected outcome based on app conten
| USK (Germany) | 0 | | USK (Germany) | 0 |
| Australia | G | | Australia | G |
Questionnaire answers (all "No" since no violence / sexual / drugs / gambling / etc. content; app is a dev tool): Questionnaire answers (all "No" — no violence / sexual / drugs / gambling content; app is a dev tool):
- Does it contain violence? No - Does it contain violence? No
- Does it contain sexual content? No - Does it contain sexual content? No
- Does it contain crude humor? No - Does it contain crude humor? No
@@ -230,7 +273,7 @@ If reviewers cannot self-host, contact support@vibebrowser.app and we will provi
## Release notes / "What's new" per release ## Release notes / "What's new" per release
Already wired in CI: `distribution/whatsnew/whatsnew-en-US`. Wired in CI: `distribution/whatsnew/whatsnew-en-US`.
Bump this file before tagging a release. Keep it under 500 chars. Per-language variants supported (`whatsnew-fr-FR`, `whatsnew-de-DE`, etc.). Bump this file before tagging a release. Keep it under 500 chars. Per-language variants supported (`whatsnew-fr-FR`, `whatsnew-de-DE`, etc.).
@@ -243,18 +286,18 @@ Bump this file before tagging a release. Keep it under 500 chars. Per-language v
3. **Open testing** — optional intermediate step. 3. **Open testing** — optional intermediate step.
4. **Production** — only after Closed testing requirements met. 4. **Production** — only after Closed testing requirements met.
CI currently publishes to `internal` track ✅ correct for first release. CI currently publishes to `internal` track. ✅
--- ---
## Pending before first publish ## Pending before first publish
- [ ] Identity verification (governor ID upload, Google review days) - [ ] Identity verification (government ID upload, Google review)
- [ ] App icon (real PNG, not placeholder) - [ ] App icon (real PNG, not placeholder)
- [ ] Adaptive icon (real PNG) - [ ] Adaptive icon (real PNG)
- [ ] Feature graphic 1024×500 - [ ] Feature graphic 1024×500
- [ ] At least 2 phone screenshots - [ ] At least 2 phone screenshots
- [ ] Privacy policy live at https://opencode.vibebrowser.app/privacy - [ ] Privacy policy live at https://opencode.vibebrowser.app/privacy
- [ ] Decide pricing model (see monetization research) - [ ] Decide pricing model
- [ ] Run IARC content rating questionnaire (after app created) - [ ] Run IARC content rating questionnaire (after app created in Play Console)
- [ ] Complete Data safety form (after app created) - [ ] Complete Data safety form (after app created in Play Console)

8
distribution/whatsnew/whatsnew-en-US
View File

@@ -1,7 +1,3 @@
v0.4.1 — Stability and UX improvements. v0.4 — streaming chat, diff viewer, tool-call approval, biometric unlock.
- Improved markdown rendering with copy-to-clipboard for code blocks Watch your AI coding agent think token by token. Review every file diff before it lands. Approve or reject tool calls (file writes, shell commands) from your phone. Biometric lock keeps sessions private. Connect via local Wi-Fi, Cloudflare Tunnel, ngrok, or Tailscale.
- Better session status indicators (idle/busy/retry)
- Connection management improvements
- Biometric authentication gate
- Bug fixes and performance improvements