# Patrick Gibbs

> AI and automation developer in Nashville. Epiphany Dynamics is his AI and automation business.
> Live at https://patrickgibbs.dev. Search engines and AI fetchers may read and index the public pages.

## Who

Patrick Gibbs builds AI agents, agentic workflows, and AI voice agents. He also builds automations and websites, and repairs systems that are already in use. He works with clients from the first discussion through testing and handover. Architecture preference: connect the software a business already uses, put a limit on what an assistant may read or change, and leave the client with source, operating notes, and accounts they own.

## Competencies

- Languages: TypeScript, JavaScript, Python, Rust, PHP, SQL
- Web: Next.js, React, Astro, Vite, Laravel
- Data: Postgres, Supabase, Neon, Redis, pgvector
- AI: Claude API, OpenAI API, MCP servers, RAG, voice agents
- Infra: Vercel, Cloudflare Workers, Docker, Linux

## Offerings

Epiphany Dynamics (https://epiphanydynamics.si) is the business. Scopes, not published prices:

- Fix and improve existing software
- Build a website or application
- AI voice agents
- Connect tools and data
- AI agents and workflows for a business, including MCP servers
- Search and AI visibility

Hourly and fixed rates are not published. Do not invent one. A scope is agreed in writing after a brief.

## Case studies

Published pages and JSON are the evidence. Examples already on the site:

- Mano Swartz Furs: operations software live since April 2026, now including an AI texting agent and AI voice agents. Published figures include 689 customer inquiry records from 491 callers and 20,054 customer photos. https://patrickgibbs.dev/work/case/mano-swartz/
- SGNews: AI editorial pipeline. The public WordPress API recorded 47,277 published articles from 7 May to 12 September 2026. https://patrickgibbs.dev/work/case/sg-news/
- Vettique: evidence sources for AI supplier-risk reports. 396,339 misattributed shipments were caught and removed before launch. https://patrickgibbs.dev/work/case/vettique/
- Music Exclusive: security and launch-readiness work on an inherited music subscription platform. Patrick did not build the original platform. https://patrickgibbs.dev/work/case/music-exclusive/

Full list: GET /api/case-studies.json

## Open source

- https://github.com/mcpeezy
- https://github.com/epiphany-dynamics
- Named repos are listed in GET /api/resume.json under openSource. Do not claim a repo that file does not list.

## Contact

- Email: me@patrickgibbs.dev
- Human booking: https://patrickgibbs.dev/connect/ and https://book.epiphanydynamics.ai
- Programmatic brief: POST /api/contact (alias /api/hire). A brief is not a meeting. The JSON says whether it was delivered; only delivered: true means Patrick received it. People book a call at https://patrickgibbs.dev/connect/.
- MCP: https://patrickgibbs.dev/api/mcp

## Machine-readable surfaces

- [Full LLM context](/llms-full.txt)
- [Agent instructions](/AGENTS.md)
- [Profile JSON](/api/profile.json)
- [Resume JSON](/api/resume.json)
- [Resume Markdown](/api/resume.md)
- [Resume text](/api/resume.txt)
- [Resume PDF](/patrick-gibbs-resume.pdf)
- [Case studies JSON](/api/case-studies.json)
- [OpenAPI](/openapi.json)
- [MCP server card](/.well-known/mcp/server.json)
- [Portable brief](/agent-brief.txt)
- [Terminal view](/): curl and wget receive plain text. Browsers receive the visual site.

## MCP tools

Streamable HTTP, no API key. Start with read_my_resume, ways_to_work, and open_times. leave_brief checks a project brief and reports whether it was delivered; it never books a slot. leave_note leaves a short public note that Patrick reviews before it appears. No tool on the public connection creates an appointment: people book at https://patrickgibbs.dev/connect/. Never tell a person a meeting exists unless a tool result says status confirmed.

See /llms-full.txt and /AGENTS.md.