Documentation Index
Fetch the complete documentation index at: https://panopticon-cli.com/llms.txt
Use this file to discover all available pages before exploring further.
Panopticon CLI
“The Panopticon had six sides, one for each of the Founders of Gallifrey…”
— Classic Doctor Who. The Panopticon was the great hall at the heart of the Time Lord Citadel, where all could be observed. We liked the metaphor.
No install step required — npx @panctl/cli starts the browser/server experience immediately. Missing tools are prompted and installed inline the first time you use a feature that needs them.
Panopticon is an open-source control plane for multi-agent software development. It connects to your issue tracker — GitHub Issues, Linear, GitLab, or Rally — and manages the full agent lifecycle from planning through merge.
Assign an issue to Panopticon and an Opus-class model reads the ticket and writes a detailed PRD, then hands off to a cost-effective implementation model (configurable per project — Kimi, Sonnet, GPT, Gemini, or others) that writes code in an isolated git worktree. Once the implementation agent signals completion, five specialist agents take over: an inspect specialist verifies each implementation step against the spec, a review specialist diffs the changes against the PRD, a test specialist runs your CI pipeline, a UAT specialist performs browser-based requirement verification, and a merge specialist opens the PR. The only human step is clicking Merge.
Each stage runs as a separate agent process in its own tmux session with full terminal I/O — attach to any session mid-flight, read its scrollback, or send it commands directly. Coordinating all of this is Cloister, a lifecycle manager that routes tasks to models based on capability and cost, monitors agent health, detects and recovers stuck sessions, tracks per-issue token spend, and orchestrates the handoff sequence between specialists. Everything is observable through Mission Control, a 13-view real-time dashboard.
Architecture at a Glance
Panopticon started as a CLI for orchestrating coding agents and grew into Command Deck, a desktop app. The CLI, the GUI, and any script that can make an HTTP request all drive the same REST surface — so you can spawn an agent from a kanban card, a terminal, or a webhook without switching tools. Under the hood: an Effect.js + TypeScript server, a React frontend over typed WebSocket RPC, SQLite for state, and Electron as the shell. Launch the browser/server experience with npx @panctl/cli; use @panctl/desktop for the packaged desktop app, and keep pan for headless and CI.
Why Panopticon?
- Stop babysitting agents. Spawn them from a dashboard, monitor progress in real time, and let specialists handle code review, testing, and merging.
- Use the right model for the job. Opus for planning, Kimi for implementation, Haiku for quick commands — automatic routing based on task type and capabilities.
- Work survives across sessions. PRDs, state files, beads, and skills persist context so agents don’t start from zero every time.
- One skill format, every tool. Write a SKILL.md once and it works across Claude Code, Codex, Cursor, and Gemini CLI.
How It Works
Plus two specialists that run inline: Inspect verifies each implementation step against the spec during development, and UAT performs browser-based requirement verification after tests pass.
Key Features
| Feature | Description |
|---|
| Multi-Agent Orchestration | Spawn and manage AI agents in tmux sessions via dashboard or CLI |
| Cloister Lifecycle Manager | Automatic model routing, stuck detection, cost tracking, and specialist handoffs |
| Mission Control | 13-view dashboard — project tree, activity feed, kanban board, agent status, costs, metrics, and more |
| PRD-Driven Workflow | Opus writes a PRD before implementation starts; agents are blocked without one |
| 70+ Universal Skills | Pre-built skills ship out of the box, synced via pan sync — one SKILL.md works across all AI tools |
| Multi-Tracker Support | GitHub Issues, Linear, GitLab, Rally — all from one dashboard |
| Multi-Model Routing | Anthropic, OpenAI, Google, Kimi, Zhipu — route by task type, capability, and budget |
| Workspaces | Git worktree-based feature branches with Docker isolation (local and remote via Fly.io) |
| Convoys | Run parallel agents on related issues with automatic synthesis |
| 5 Specialist Agents | Review, test, inspect, UAT, and merge — fully automated quality pipeline |
| Beads | Git-backed task tracking that survives context compaction and works offline |
| Cost Tracking | Per-issue, per-stage token costs with dashboard analytics |
| Legacy Codebase Support | AI self-monitoring skills that learn your codebase conventions over time |
| Tool | Support |
|---|
| Claude Code | Full support — agent runtime, hooks, skills |
| Codex | Skills sync |
| Cursor | Skills sync |
| Gemini CLI | Skills sync |
| Google Antigravity | Skills sync |
Dashboard Views
The dashboard at https://pan.localhost provides 13 views:
| View | Purpose |
|---|
| Mission Control | Project tree + activity timeline — see the full pipeline for any feature |
| Board | Kanban board with cost badges, agent status, and workspace controls |
| Agents | Cloister Deacon, specialist agents, and issue agents with token/cost tracking |
| Resources | System resource monitoring and allocation |
| Convoys | Parallel agent runs with synthesis status |
| Handoffs | Specialist handoff queue and history |
| Activity | Real-time agent command output log |
| Metrics | Runtime comparison and performance analytics |
| Costs | Per-issue, per-stage cost breakdown with daily totals |
| Skills | All available skills with descriptions and sync status |
| Health | System health checks and diagnostics |
| God View | Aggregate cross-project view of all agent activity |
| Settings | Model routing, tracker API keys, and project configuration |
Quick Start
npx @panctl/cli starts the browser/server experience immediately. For the packaged desktop app, use @panctl/desktop. For headless and CI, use pan (install via npm install -g @panctl/cli).
Dashboard runs at https://pan.localhost (or http://localhost:3011 if you skip HTTPS setup).
Learn More
