# askbowtie

> Every signal a website emits (analytics, JS errors, conversions, Google Search Console, Google Ads, uptime) joined into one MCP server an AI agent can query. Ask in plain language, get answers with every number cited. No dashboards.

askbowtie is built for website owners, vibecoders, and agencies who work with AI agents
(Claude Code, Codex, Cursor, Claude Desktop, or anything that speaks MCP). One lightweight
tracker plus connected Google sources feed one per-site database; answers are joined across
silos at read time (ads ⋈ analytics ⋈ search ⋈ errors ⋈ uptime). "Bowtie" is the persona you
address; the ⋈ mark is the relational-algebra bowtie (join) operator.

Key facts:
- Primary interface: an MCP server your agent connects to once. Tools cover traffic,
  conversions, incidents, alerts, performance, page-level detail, and cross-silo joins.
- Human path: email digests and down/up alerts. No dashboard required, by design.
- Data: first-party only, stored per site in its own database, exportable, never sold or shared.
- Tracker: one small async script; never blocks rendering.
- Multi-site: each site gets its own tracker and database under one connection.

## Pages

- [Home](https://askbowtie.com/): Ask Bowtie about your site — interactive demo, features, FAQ
- [Features](https://askbowtie.com/#features): the six signals, the join, MCP server, email digest
- [FAQ](https://askbowtie.com/#faq): data collected, tracker weight, MCP setup, supported agents, privacy, pricing

## Docs

- [Docs](https://askbowtie.com/docs/): tracking reference — install, auto-tracked events, custom events, conversions with revenue, errors, guardrails, MCP setup
- [events.md](https://askbowtie.com/docs/events.md): the same reference in plain markdown, written for agents

## Full detail

- [llms-full.txt](https://askbowtie.com/llms-full.txt): complete product description, MCP tool list, and FAQ text