kensaurus Mushi Mushi — AI QA Platform Setup & Usage
Mushi Mushi is an AI-powered quality engineering platform that provides:
- Bug Intelligence — capture and triage reports from real users
- Story Mapping — automatically discover user stories from a live URL (no YAML required)
- TDD Generation — turn user stories into Playwright test specs with one command
- PDCA Auto-improve — failing tests are automatically rewritten by Claude
- Multi-key BYOK — quota-aware API key pool with automatic failover
Quick Start
1. Install the SDK
npm install @mushi-mushi/web
import { Mushi } from '@mushi-mushi/web'
Mushi.init({
projectId: 'YOUR_PROJECT_ID',
apiKey: process.env.MUSHI_API_KEY!,
capture: {
discoverInventory: { enabled: true },
},
})
2. Install the CLI
npm install -g @mushi-mushi/cli
mushi login --api-key mushi_... --endpoint https://<ref>.supabase.co/functions/v1/api --project-id <pid>
3. Map user stories from a live URL (no YAML needed)
mushi stories map --url https://your-app.vercel.app --wait
Or in the console: Inventory → Discovery → "Map from live app"
4. Generate TDD tests from user stories
# List discovered stories first (from accepted inventory)
# Then generate a test for a specific story
mushi tdd gen login-flow --mode review
# Approve/reject the generated test
mushi tdd pending
mushi tdd approve <qa-story-id>
5. Auto-improve failing tests (PDCA)
mushi tdd improve
This analyzes recent failures and rewrites the test scripts with Claude. New versions are created with source=pdca and gated by the story's automation_mode.
6. Manage API key pool (BYOK)
# List all keys and their status
mushi keys list
# Add a backup key (used automatically on quota exhaustion)
mushi keys add --provider anthropic --key sk-ant-... --label "backup" --priority 200
MCP Integration (for Cursor / Claude Desktop)
Add to your cursor.mcp.json or mcp.json:
{
"mcpServers": {
"mushi": {
"command": "npx",
"args": ["@mushi-mushi/mcp"],
"env": {
"MUSHI_API_KEY": "mushi_...",
"MUSHI_API_ENDPOINT": "https://<ref>.supabase.co/functions/v1/api",
"MUSHI_PROJECT_ID": "<project-id>"
}
}
}
}
Available MCP tools:
| Tool | What it does |
|---|---|
map_user_stories |
Crawl a live URL, draft user stories with Claude |
generate_tdd_from_story |
Write a Playwright test from a story id |
improve_qa_story |
PDCA auto-improve failing tests |
run_qa_story |
Trigger a manual QA run |
list_pending_review_stories |
See what TDD tests need approval |
approve_qa_story |
Approve or reject a pending test |
list_byok_keys |
See API key pool health |
add_byok_key |
Add a backup key |
get_recent_reports |
Survey the bug triage queue |
dispatch_fix |
Auto-fix a report with a Cursor Cloud agent |
Console Pages
| Page | URL | What to use it for |
|---|---|---|
| Inventory | /inventory |
Accept user story proposals, review map runs |
| QA Coverage | /qa-coverage |
Review pending TDD tests, see test history |
| Settings → API Keys | /settings/llm-keys |
Manage BYOK key pool |
| Reports | /reports |
Bug triage queue |
Automation Modes
When generating TDD tests, set --mode:
| Mode | Behaviour |
|---|---|
auto |
Test is immediately enabled in the QA schedule |
review |
Test waits in the "pending review" queue |
approve |
Test is created disabled; you manually enable |
BYOK Key Pool
Mushi uses a priority-ordered pool of API keys per provider. When a key hits its quota:
- Mushi marks it
quota_exhaustedwith a 1-hour cooldown - Automatically falls back to the next active key
- Shows a banner in the console with a "Switch key" CTA
Add multiple keys per provider to avoid interruption:
mushi keys add --provider anthropic --key sk-ant-primary --label "primary" --priority 100
mushi keys add --provider anthropic --key sk-ant-backup --label "backup" --priority 200
Environment Variables
| Variable | Description |
|---|---|
MUSHI_API_KEY |
SDK / CLI API key (prefix: mushi_) |
MUSHI_API_ENDPOINT |
Full URL to the Mushi API edge function |
MUSHI_PROJECT_ID |
Project UUID from the console |