Event Horizon

The multi-agent coordination framework for AI coding. Give your AI agents a shared plan, file locks, and real-time awareness of each other — so they build together instead of breaking each other's work.

The Problem

You open three terminals. Claude Code is building the API. OpenCode is writing tests. Copilot is updating the docs. They're all editing the same project.

Then Claude overwrites OpenCode's changes to server.ts. OpenCode doesn't know. The build breaks. You spend 20 minutes untangling the mess.

Event Horizon prevents this entirely.

How It Works

1. Create a plan

/eh:create-plan Build a REST API with auth, database layer, and tests

Event Horizon generates a plan with parallel tracks, dependency annotations, and verify steps. Each task is concrete — file paths, function signatures, expected behavior.

2. Assign agents

Terminal 1:  /eh:work-on-plan api Phase 1 — database
Terminal 2:  /eh:work-on-plan api Phase 2 — endpoints
Terminal 3:  /eh:work-on-plan api Phase 3 — tests

Each agent claims its tasks atomically. No two agents can claim the same task. Dependencies are enforced — Phase 2 can't start until Phase 1 is done.

3. Watch them coordinate

The Kanban board updates live as agents move through tasks. The Universe view shows planets, orbital debris for plan tasks, ships for data transfers, and lightning when agents touch the same file.

If two agents try to edit the same file, the second one is hard-blocked — not warned, blocked. It sees who holds the lock, works on something else, and retries when the lock releases.

Features

Multi-Agent Plan Coordination

• Plan parser — any markdown checklist (- [ ]) with numbered IDs and - depends: annotations
• Atomic task claiming — no race conditions, no duplicate work
• Dependency resolution — blocked tasks auto-unblock when dependencies complete
• Agent messaging — targeted or broadcast messages between agents
• Auto-discovery — new agents are notified about active plans on spawn
• Multi-plan support — load multiple plans, manage lifecycle (active/completed/archived)
• Checkbox sync — completed tasks update the source markdown file automatically

File Locking

When Agent A writes to a file, Agent B is hard-blocked from accessing it. Locks auto-expire after 30 seconds, refresh on each write, and release on agent termination. No manual lock management needed.

Live Visualization

Each AI agent is a planet. The cosmic metaphor encodes real information:

Operations Dashboard

Full-screen dashboard alternative (Ctrl+Shift+E O) with:

• Agents / Plans sidebar — agents grouped by workspace, plans grouped by status (active/completed/archived)
• Overview — metrics grid, tool breakdown charts, agent summary table
• Files — sortable heatmap of file activity across all agents
• Logs — searchable event log with type filters
• Timeline — horizontal swimlane of agent activity over time
• Plan — Kanban board with sticky headers, progress bar, column toggle

Skills Management

Discover, create, browse, duplicate, and organize Agent Skills. Three coordination skills ship with the extension:

15 MCP Coordination Tools

All agents access these tools via the MCP server (auto-registered on connect):

eh_load_plan eh_get_plan eh_list_plans eh_claim_task eh_update_task eh_archive_plan eh_delete_plan eh_send_message eh_get_messages eh_check_lock eh_acquire_lock eh_release_lock eh_wait_for_unlock eh_list_agents eh_file_activity

Achievements

28 medals tracking milestones — from first agent spawn to multi-agent file collisions, UFO encounters, and astronaut black hole dives. Some are secret. Some have tiers.

Supported Agents

One-click connect for Claude Code, OpenCode, and Copilot. MCP server auto-registered in each agent's config. Hooks auto-updated on every extension activation — no manual reinstall needed.

Privacy

• 100% local — HTTP server on 127.0.0.1:28765. Nothing leaves your machine.
• Zero agent overhead — hooks use --connect-timeout 2 with silent fallback. If Event Horizon is closed, agents run identically.
• No telemetry — no analytics, no tracking, no data collection.

The Origin

I asked Claude:

"If you could choose a visual representation of yourself as an AI agent, how would you represent yourself and your fellow AI agents collaborating?"

Claude's answer:

"Each agent is a planet — a massive entity that consumes energy, emits output, and exerts gravitational influence. Tasks orbit as moons. Data flows as ships. At the center, a black hole where completed work collapses. This scales naturally. One agent is a lonely planet. Five agents become a solar system."

From that answer, Event Horizon was born.

Getting Started

Install from the VS Code Marketplace, then:

1. Open: Rocket icon in editor title bar, or Ctrl+Shift+E H
2. Connect: Click Connect → choose agent → Install
3. Code: Launch an agent session. Planet appears.

No agent? Click Demo for simulated agents.

Development

See CONTRIBUTING.md for full guidelines.

pnpm install && pnpm build    # build all packages
pnpm test                      # run tests (500+)
pnpm dev                       # watch mode

Press F5 to launch the Extension Development Host.

Documentation

• Changelog
• Architecture
• Contributing
• Code of Conduct

License

MIT License with Commons Clause — see LICENSE.