Healix
End-to-end AI test generation and QA automation server that orchestrates auto-detection, browser exploration, test generation, execution, and dashboard analysis directly from your IDE.
README
Healix
End-to-end AI test generation and QA automation, from your IDE.
Healix is a monorepo containing two packages:
testbot-mcp/—@healix/mcp(v2.0.0), an MCP server installed in developer IDEs. It orchestrates the full testing pipeline: auto-detect → explore → generate → execute → analyze → dashboard.webapp/— A Next.js 16 / React 19 webapp deployed on Vercel. Hosts all AI calls (OpenAI server-side), the Supabase-backed dashboard, artifact storage, and the Inngest async generation pipeline.
Users only need a HEALIX_API_KEY. All AI calls are proxied server-side through the webapp — no OpenAI key needed on user machines.
Quick Start
1. Install the MCP server
npm install -g @healix/mcp
2. Configure your IDE
Add to your MCP settings (~/.cursor/mcp.json, Claude Code, or Windsurf):
{
"mcpServers": {
"healix": {
"command": "npx",
"args": ["@healix/mcp"],
"env": {
"HEALIX_API_KEY": "your-healix-api-key"
}
}
}
}
Get your API key from the Healix dashboard → API Keys.
3. Run a test
In your IDE prompt:
"Test my app using the healix mcp"
The pipeline runs automatically:
- Config UI — local HTTP form collects test type, base URL, start command, PRD file (optional), and role credentials
- Auto-detect — port, framework, and start command inferred from the project when not supplied
- Browser exploration — browser-use (Python subprocess) discovers real flows; falls back to a zero-dependency Playwright heuristic explorer if browser-use is unavailable
- PRD/AC parse — structured acceptance-criteria extraction via
/api/parse-prd; AC-less apps use exploration flows directly - Test generation — GPT-powered multi-agent fan-out via
/api/generate-tests; each test tagged[REQ:F#.S#.AC#] - Credential injection — per-role Playwright
storageStatewritten to.healix/auth-state-{role}.json - Tiered execution — three Playwright projects run in sequence:
- Tier A (
tierA-public) — unauthenticated / public flows - Tier B (
tierB-auth-{role}) — one project per role; markedblockedif login fails (A + C still run) - Tier C (
tierC-backend) — API/backend contract tests
- Tier A (
- Artifact upload — screenshots, videos, and Playwright traces uploaded to Supabase Storage
- Ingest — results POSTed to
/api/test-runs/ingest - Dashboard deep-link — opens the run page with tier pills, failure analysis, and artifacts
MCP Tools
| Tool | Description |
|---|---|
healix_test_my_app |
Full end-to-end pipeline. Accepts projectPath, baseURL, port, startCommand, testType (frontend/backend/both), prdFile, generateTests, openDashboard, credentials, codebaseContext, playwrightMcp, and more. |
healix_configure |
Opens the config UI and returns validated settings without running the pipeline. Useful for pre-flight checks. |
Configuration
MCP environment variables
Set these in the env block of your MCP client config (Cursor, Claude Code, Windsurf):
| Variable | Required | Description | Default |
|---|---|---|---|
HEALIX_API_KEY |
Yes | Authenticates MCP → webapp; meters token usage. | — |
HEALIX_DASHBOARD_URL |
No | Webapp base URL. Used for all API calls and dashboard deep-links. | Production Vercel URL |
HEALIX_RUN_BUDGET_MS |
No | Overall pipeline timeout (ms). | 7200000 (120 min) |
HEALIX_GEN_BUDGET_MS |
No | Test-generation stage timeout (ms). Raise for large codebases; otherwise Healix expands it for large/xlarge discovered apps. | 1800000 (30 min) |
HEALIX_GENERATION_AGENT_CONCURRENCY |
No | Number of generation agents to run at once. Lower for fragile local webapps, raise for stable production webapps. | 3 |
HEALIX_GENERATION_AGENT_TIMEOUT_MS |
No | Explicit per-agent generation transport timeout. By default Healix derives this from the remaining generation budget and codebase complexity. | derived |
HEALIX_SKIP_PLANNER |
No | Set 1 to bypass the pre-fan-out planner pass (emergency circuit breaker). |
unset |
Webapp environment variables
Copy .env.example to webapp/.env.local:
| Variable | Description |
|---|---|
DATABASE_URL |
PostgreSQL connection string (Supabase pooler URI for production). |
NEXT_PUBLIC_SUPABASE_URL |
Supabase project URL. |
NEXT_PUBLIC_SUPABASE_ANON_KEY |
Supabase anon key (public). |
SUPABASE_SERVICE_ROLE_KEY |
Supabase service role key (server-side only — never expose). |
OPENAI_API_KEY |
OpenAI API key. Used only by webapp API routes — never by the MCP. |
OPENAI_MODEL |
Model override (default: gpt-5.5-mini). |
HEALIX_GEN_ASYNC |
true routes /api/generate-tests through Inngest background jobs. Default: false. |
INNGEST_EVENT_KEY |
Inngest event key (required only when HEALIX_GEN_ASYNC=true). |
INNGEST_SIGNING_KEY |
Inngest webhook signing key (required only when HEALIX_GEN_ASYNC=true). |
INNGEST_DEV |
Set 1 when running against the local inngest-cli. |
HEALIX_PLANNER_AGENT |
1 to enable LLM-driven planner. Default: 0 (rule-based). |
AI_MAX_REQUESTS_PER_MINUTE |
Per-user AI rate-limit ceiling. Default: 20. |
See .env.example for the full annotated reference.
Async generation (Inngest)
The default sync generation path runs the full multi-agent fan-out inside the /api/generate-tests request. For very large codebases where a single agent call risks hitting Vercel's function timeout:
- Set
HEALIX_GEN_ASYNC=trueon the webapp and configureINNGEST_EVENT_KEY+INNGEST_SIGNING_KEY. - No MCP change needed —
@healix/mcpauto-detects the202response and polls/api/generate-tests/jobs/{jobId}.
Do not enable HEALIX_GEN_ASYNC=true for local dev. There is no function-timeout constraint on localhost; the async path adds Inngest setup and polling overhead for no benefit.
Development
# Start the webapp (Next.js dev server on port 3000)
npm run dev:webapp
# Start the MCP server (stdio transport)
npm run start:testbot
# Run MCP unit tests (26 test files via node --test)
npm run test:testbot
# Database — run from webapp/
npm run db:generate # generate Drizzle migration from schema changes
npm run db:migrate # apply pending migrations
npm run db:studio # open Drizzle Studio
Architecture
User IDE (Cursor / Claude Code / Windsurf)
│ stdio (MCP protocol)
▼
@healix/mcp ─── thin client, HEALIX_API_KEY only ───────────────┐
│ │
│ Pipeline (pipeline-worker.js): │
│ 1. Config UI (local HTTP server) │
│ 2. Auto-detect port/framework/start-cmd │
│ 3. Browser-use exploration (Python) or PW heuristic │
│ 4. Parse PRD/AC ───── HTTPS ──────────────────────────┤
│ 5. Generate tests ──── HTTPS ──────────────────────────┤
│ 6. Inject credentials (storageState per role) │
│ 7. Run Playwright: Tier A / Tier B / Tier C │
│ 8. Upload artifacts ─── HTTPS ─────────────────────────┤
│ 9. Ingest results ──── HTTPS ──────────────────────────┤
│ 10. Open dashboard deep-link │
│ ▼
└──────────────────── Healix webapp (Vercel / localhost) ──
│ Next.js 16 + React 19
│ Supabase (Auth + DB + Storage)
│ Drizzle ORM (PostgreSQL)
│ Inngest (async generation jobs)
└──────── OpenAI (server-side only)
Project Structure
TestBot_MCP/
├── testbot-mcp/ # @healix/mcp — npm package (v2.0.0)
│ ├── bin/healix-mcp.js # CLI entry point
│ ├── src/
│ │ ├── index.js # MCP server + tool registration
│ │ ├── pipeline-worker.js # End-to-end pipeline orchestration
│ │ ├── auto-detector.js # Port/framework/start-cmd detection
│ │ ├── webapp-client.js # All webapp API calls (HEALIX_API_KEY)
│ │ ├── config-ui-launcher.js # Local HTTP config form
│ │ ├── browser-use-driver.js # Python browser-use subprocess
│ │ ├── playwright-explorer.js # Zero-dep Playwright heuristic fallback
│ │ ├── playwright-integration.js # Tier A/B/C Playwright projects
│ │ ├── playwright-mcp-client.js # @playwright/mcp integration
│ │ ├── credentials-injector.js # Per-role storageState injection
│ │ ├── artifact-uploader.js # Supabase Storage upload
│ │ ├── results-merger.js # Merge tier results + blocked status
│ │ ├── report-generator.js # Build ingest payload
│ │ ├── mcp-telemetry.js # Background telemetry reporting
│ │ ├── multi-service-starter.js # App-under-test launcher
│ │ ├── context-gatherer.js # Codebase context extraction
│ │ ├── dashboard-launcher.js # Open dashboard deep-link
│ │ ├── logger.js
│ │ ├── ai-providers/
│ │ │ └── saas-client.js # Proxy → Healix webapp
│ │ └── failure-triage/
│ │ ├── classifier.js # Deterministic first-match rules
│ │ ├── agent-response.js # AI failure analysis parsing
│ │ ├── error-remediations.js # Patch suggestions
│ │ ├── evidence-bundler.js # Test source + AC + trace evidence
│ │ ├── pipeline-error-classifier.js
│ │ └── trace-parser.js # Playwright trace.zip parser
│ ├── scripts/
│ │ ├── browser_use_runner.py # Pinned browser-use driver
│ │ └── localhost-smoke.js # Local smoke test helper
│ └── test/ # 26 test files (node --test)
│
└── webapp/ # Next.js 16 webapp (Vercel)
├── src/
│ ├── app/
│ │ ├── (auth)/ # Sign in / sign up / callback pages
│ │ ├── (dashboard)/ # Authenticated dashboard pages
│ │ │ ├── home/ # Overview + recent runs
│ │ │ ├── create-tests/ # Manual test run creation
│ │ │ ├── all-tests/ # Test run history
│ │ │ ├── mcp-tests/ # MCP-originated runs
│ │ │ ├── test-run/[id] # Live + historical run detail
│ │ │ ├── import-tests/ # Groovy test import from Excel/CSV
│ │ │ ├── monitoring/ # Live MCP telemetry + run monitoring
│ │ │ ├── api-keys/ # API key management
│ │ │ ├── plan-billing/ # Plan + credits
│ │ │ └── profile/
│ │ └── api/
│ │ ├── generate-tests/ # Multi-agent AC-traced generation
│ │ ├── parse-prd/ # PRD → structured AC extraction
│ │ ├── exploration/plan/ # Flow prioritization
│ │ ├── analyze-failures/ # AI failure triage
│ │ ├── test-runs/ # CRUD + ingest + phase + SSE stream
│ │ ├── import-tests/ # Excel/CSV → Groovy generation
│ │ ├── mcp-telemetry/ # Telemetry ingest + summary
│ │ ├── mcp-auth/validate/ # API key validation
│ │ ├── artifacts/ # Supabase Storage proxy
│ │ ├── upload-artifacts/ # Signed upload URLs
│ │ ├── api-keys/ # Key CRUD
│ │ ├── auth/ # Supabase Auth helpers
│ │ ├── inngest/ # Inngest function registry
│ │ └── profile/
│ └── lib/
│ ├── db/schema.ts # Drizzle ORM schema (source of truth)
│ ├── test-generation/ # GPT orchestration: planner + agents
│ ├── inngest/functions/ # generate-tests-orchestrator + agent
│ ├── types/database.ts # Shared TypeScript interfaces
│ ├── ai-guard.ts # Per-user AI rate limiting
│ ├── credits.ts # Token accounting
│ ├── mcp-live-runs.ts # SSE live run state from telemetry
│ ├── groovy-generator.ts # Groovy test file generation
│ └── excel-parser.ts # Excel/CSV test case ingestion
└── drizzle/ # SQL migrations
Database Schema (key tables)
| Table | Purpose |
|---|---|
profiles |
User accounts (plan, credits, tokens) |
api_keys |
Hashed API keys per user |
test_runs |
Execution results with tier results, pipeline errors, AI analysis |
test_failures |
Per-test failure verdicts with evidence and user overrides |
mcp_telemetry_events |
Pipeline event stream (powers live run monitoring) |
generation_jobs |
Inngest async generation job lifecycle |
generation_plans |
24h cached planner output per repo snapshot |
test_artifacts |
Screenshot / video / trace metadata + Supabase Storage paths |
import_sessions |
Groovy import sessions (Excel/CSV → test cases) |
imported_test_cases |
Parsed test cases from import |
generated_groovy_files |
Generated Groovy API test files |
Migrating from @testbot/mcp
See MIGRATION.md for the full upgrade guide. Summary:
npm uninstall -g @testbot/mcp
npm install -g @healix/mcp
Remove OPENAI_API_KEY, AI_PROVIDER, and any other AI keys from your MCP config — they are ignored in v2.0.0. Only HEALIX_API_KEY is required.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.