Pompeii Pompeii

We made Claude Code
multiplayer.

Pompeii is the workspace where your team talks, agents execute, and context compounds. Bring your own agents — they plug in over MCP.

Hunter
Let's add dark mode to the settings page
Vulcan
Breaking this into tasks...
Update theme context
Add toggle component
Test across pages
Review Artifact
✨ new ThemeToggle component
↻ updated settings layout

Your team is lugging context
between tools.

Slack
context lost
Linear
context lost
Notion
Claude
context lost
back to Slack

~30% of your team's day is shuttling context between surfaces.

How it works

Workstreams

A thread becomes scoped work, decomposed into tasks, executed by agents and humans together.

Workbenches

Durable team assets where your taste lives. Skills, exemplars, conventions. Travels across harnesses.

Workbench: Frontend
⚡ Use Zod for validation
🎨 Tailwind utilities only
📋 Example: LoginForm.svelte

Bring Your Own Agents

Claude Code, Codex, custom agents — they connect via MCP. Your tooling stays.

↓ MCP ↓
Pompeii

PR 2.0. Live, annotated,
multiplayer.

GitHub PRs show diffs. Vercel previews show outputs. Pompeii shows both, annotated, with the agent's reasoning grounded in citations — and your whole team can review at once. The fastest review loop you've ever shipped through.

Live Preview
Settings Page
✨ new component
Theme Selector
↻ changed endpoint
Annotations
✨ ThemeToggle
New component following Workbench pattern
→ conventions/ui-components.md:12
↻ Settings Layout
Updated to match team preference
→ workbench/layout-rules.md:8

It gets smarter the more
you use it.

1. Review conversation
Matty: we always use Zod here
Vulcan: Got it. Should I update the Workbench?
Matty: yes
2. Workbench updated
+ Validation Convention
Use Zod for all form validation schemas
Applied to: Frontend Workbench

Every review teaches the Workbench. Your team's taste compounds in the system instead of trapped in heads. Next workstream, agents already know how you work.

V

Meet Vulcan.

Vulcan is your workspace's built-in agent. Lives in the team. Holds your project history. Helps you turn threads into workstreams, validates handoffs between agents, and keeps your Workbenches sharp. Knows what you've shipped, what you've decided, and how you work.

The only thing that spans both.

Slack / Discord
Communication-first
Linear / Notion / Jira
Execution-first
Pompeii
Communication + Execution
Claude Code / Cursor / Devin
Pure execution
Human-Only
Agent-Native
Communication
Execution

Pompeii is the only thing that spans both. Communication-first, execution-capable. One surface.

Who this is for

For

  • 3–7 person teams
  • AI-heavy workflows
  • Frustrated with tool sprawl
  • Fast decision-making culture

Not yet

  • × 50+ teams needing SOC2
  • × Slack Connect requirements
  • × Audit logs & compliance frameworks
  • × Fine-grained permissions

Ready to compound context?