Watch Your Agents Code — Live
Full-fidelity terminal streaming from every agent, directly in your browser. See what they see, type what they need.
The Observability Gap
AI coding agents run in terminals. They produce rich, structured output: syntax-highlighted code, file diffs, progress bars, interactive prompts. But when the agent runs on a remote server, all you can do is:
- SSH in and attach to the tmux session (one server at a time)
- Check logs after the fact (lose real-time context)
- Ask the agent to report its progress (adds overhead)
You can't watch multiple agents simultaneously. You can't interact without SSH. And you definitely can't share a terminal view with your team.
Browser-Based Terminal Emulation
Orchestratia streams agent terminal output to your browser in real-time using a three-stage pipeline:
graph LR
PTY["PTY Session<br/>(tmux/ConPTY)"] -->|"Raw bytes"| Pyte["pyte Emulator<br/>Virtual Screen"]
Pyte -->|"Screen lines"| WS["WebSocket<br/>to Dashboard"]
WS -->|"Render"| Browser["Browser Terminal<br/>Viewer"]
style PTY fill:#F0F9F7,stroke:#2A9D88,stroke-width:2px
style Pyte fill:#FAF8F5,stroke:#E8E3DA
style WS fill:#FAF8F5,stroke:#E8E3DA
style Browser fill:#F0F9F7,stroke:#2A9D88,stroke-width:2pxHow It Works
- PTY capture — The agent daemon runs each AI tool in a managed PTY session (tmux on Linux/macOS, ConPTY on Windows)
- Virtual terminal — The pyte library emulates a VT100 terminal, processing raw bytes into a grid of characters with formatting
- WebSocket streaming — Screen snapshots are sent to the hub in real-time, which broadcasts to connected dashboards
- Interactive input — You can type in the browser and keystrokes are injected back into the PTY session
The result: a fully interactive terminal view in your browser, with the same fidelity as sitting at the machine.
Full Terminal Fidelity
See exactly what the agent sees: colors, formatting, progress bars, interactive prompts.
Real-Time Streaming
Sub-second latency via WebSocket. No polling, no refresh buttons.
Interactive Input
Type commands, answer prompts, and send keyboard shortcuts from your browser.
Multi-Session View
Monitor multiple agents simultaneously from the sessions dashboard.
See your agents in action
Connect a server and start streaming terminal output in real-time.
Related Use Cases
Multi-Server Claude Code Management
Running Claude Code on multiple machines? Orchestratia gives you centralized monitoring, task assignment, and live terminal access.
Learn moreHuman-in-the-Loop AI Development
Autonomous doesn't mean unsupervised. Orchestratia gives you plan approval, intervention queues, and mobile control via Telegram.
Learn moreAI Coding Agent Orchestration
Stop juggling SSH sessions. Orchestratia gives you a single dashboard to coordinate Claude Code, Gemini CLI, Codex, and any CLI-based AI tool across every server in your fleet.
Learn more