Funnel Analytics Agent

funnel-analytics-agent

English | [中文](README.zh-CN.md)

> Daily morning brief + real-time anomaly alerts for indie launches. Reads Vercel + OpenPanel + HyperDX + Supabase + Product Hunt, surfaces what's broken before your users tell you.

[](LICENSE) [](https://pypi.org/project/funnel-analytics-agent/) [](#)

Built by Alex Ji — solo founder shipping VibeXForge. Born from this thought:

> On Product Hunt launch day my dashboard tabs were Vercel + OpenPanel + HyperDX + Supabase + Product Hunt + GitHub. Cycling through them = 1 hour/day. Missing a 5xx for 2 hours = lost the launch.

What it does

Two modes, one agent:

Brief mode — runs every morning (cron), composes a markdown brief from all data sources, drops it in your Obsidian vault. You read it over coffee in 30 seconds.

Alert mode — runs every 10 minutes during your launch window, exits with code 2 if anything critical/alert severity is found. Your wrapper script pages you (Telegram, Slack, ntfy.sh).

Sources (v0.9, 8 total):

• Vercel — latest deployment state, deployments in last 24h, failed deploys
• Product Hunt — live vote count, comment count, momentum check on launch day
• Supabase advisor — security + performance lints; surfaces every CRITICAL individually
• OpenPanel — count of tracked events (signupcompleted, projectsubmit_completed, …) over last 24h; warns when an event drops to 0 (tracking regression)
• HyperDX — error log count over last N hours; severity climbs from info → warn → alert as count crosses 10 / 50 thresholds
• Build Quality — last 24h of pre-push reviews from build-quality-agent's local log (verdicts, BLOCK count, token spend)
• Agent Spend — cross-agent Anthropic $ MTD aggregated across ~/./usage.jsonl files
• VibeX — direct SQL pull of core launch-board metrics (new creators / submissions / cumulative plays / stage distribution); zero Claude calls. Configure via VIBEX_PROJECT_REF (or reuse SUPABASE_PROJECT_REF if it's the same project)

Install

git clone https://github.com/alex-jb/funnel-analytics-agent.git
cd funnel-analytics-agent
pip install -e .
cp .env.example .env
# fill in VERCEL_TOKEN, PH_DEV_TOKEN, etc

Usage

# One-shot brief (default)
funnel-analytics-agent

# Write brief to file (e.g. for Obsidian vault)
funnel-analytics-agent --out ~/Documents/alex-brain/morning-briefs/$(date +%Y-%m-%d).md

# Real-time alert mode — exits 2 if anything critical
funnel-analytics-agent --alert

# Limit to specific source
funnel-analytics-agent --source vercel --source producthunt

Cron setup for Product Hunt launch day

# Every 7 minutes during the 24h launch window — alert mode
7,17,27,37,47,57 * * * 1 /usr/local/bin/funnel-analytics-agent --alert || curl -d "PH alert" ntfy.sh/your-topic

# Daily brief at 7:03 AM (post-cap-reset)
3 7 * * * /usr/local/bin/funnel-analytics-agent --out ~/Documents/alex-brain/morning-briefs/$(date +\%Y-\%m-\%d).md

Design choices

• Each source fails independently. Vercel API down? Your brief still gets PH stats. The failed source is listed at the bottom of the brief; never kills the whole agent.
• No external dependencies (yet). Pure stdlib urllib — no requests, no httpx, no API client SDKs. Keeps install fast and pip-install-friendly on locked-down boxes.
• Markdown-first output. Output is meant to be human-readable in Obsidian / VS Code, not piped through a dashboard service.
• Exit codes for cron, not push notifications. v0.1-v0.3 don't ship a Telegram/Slack adapter — your cron wrapper does that. Keeps the agent's surface tiny.
• 7-day baseline auto-built from your own runs. No need to seed historical data. After 24h of cron runs the baseline starts forming; after 7 days it's solid. Bootstrap mode (no history) just skips delta calculation — never blocks output.

Roadmap

• [x] v0.1 — Vercel + Product Hunt sources · brief mode · alert mode · cron-friendly
• [x] v0.2 — OpenPanel, HyperDX, Supabase advisor sources (5 sources total, 30 tests)
• [x] v0.3 — 7-day baseline · delta_pct enrichment · severity promotion on >50% drops · 41 tests
• [x] v0.4 — Push notifier adapters (ntfy.sh / Telegram / Slack) + macOS launchd installer (54 tests)
• [x] v0.5 — Claude-summarized brief at top of every report (Haiku 4.5 default; ~$0.0008/run; falls back gracefully)
• [x] v0.8 — MCP server: query the morning brief / live alerts / per-source state from Claude Desktop

MCP server (Claude Desktop / Cursor / Zed)

Expose the brief, alert state, and individual sources as tools your AI assistant can call.

pip install 'funnel-analytics-agent[mcp]'

Then add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "funnel-analytics": {
      "command": "funnel-analytics-mcp",
      "env": {
        "VERCEL_TOKEN": "...",
        "VERCEL_PROJECT_ID": "...",
        "PH_DEV_TOKEN": "...",
        "PH_LAUNCH_SLUG": "your-launch-slug",
        "OPENPANEL_CLIENT_ID": "...",
        "OPENPANEL_CLIENT_SECRET": "...",
        "HYPERDX_API_KEY": "...",
        "SUPABASE_PERSONAL_ACCESS_TOKEN": "...",
        "SUPABASE_PROJECT_REF": "...",
        "ANTHROPIC_API_KEY": "..."
      }
    }
  }
}

Tools:

• get_brief(include_summary) — full markdown morning brief
• get_alerts() — alert-mode output: All clear or per-source severity
• get_source(name) — single source (vercel / producthunt / openpanel / hyperdx / supabase)
• usage_summary() — Anthropic token + $ totals from local usage log

License

MIT. ---

🧩 Part of the Solo Founder OS stack

A growing collection of MIT-licensed agents that share solo-founder-os as their base — cron, eval, reflexion, AnthropicClient, HITL queue, notifiers. Each agent is independently useful; together they cover the full one-person company workflow.

🌐 The whole stack is live in production at vibexforge.com.

| Agent | What it does | |---|---| | solo-founder-os | The shared base lib (cron · eval · reflexion · skill library · DGM-lite). Every other agent depends on it. | | build-quality-agent | Pre-push Claude diff reviewer + local build runner — catches CI-killing changes before they ship. | | customer-discovery-agent | Reddit pain-point scraper + Claude clustering for product validation. | | orallexa-marketing-agent | AI marketing agent — submit project once, get platform-native posts for X / Reddit / HN / Dev.to / 小红书 + 7 more. Powers vibexforge.com. | | vc-outreach-agent | Cold email drafter — investors (vc mode) or paying customers (customer mode, merged from customer-outreach in v0.9.0). HITL queue + SMTP sender. | | cost-audit-agent | Monthly bill audit across 6 providers (Vercel / Anthropic / OpenPanel / HyperDX / Supabase / GitHub Actions) with dollar-tagged waste findings. | | bilingual-content-sync-agent | EN ⇄ 中文 i18n diff + Claude translate + HITL apply. Batch API path @ 50% off. | | customer-support-agent | Triage user messages → auto-draft replies → HITL queue. Closes the L5 customer-support layer. | | payments-agent | Overdue-invoice reminder drafter. Stripe-shaped types + MockProvider fallback + hard money-safety prompt rules. |

Each agent's own row is omitted from its README. Install whichever solve real problems for you — pip install .

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: alex-jb
• Source: alex-jb/funnel-analytics-agent
• License: MIT
• Homepage: https://vibexforge.com

Install and usage instructions live in the source repository linked above.