Introduction
agentistics is the app that runs on your machine and brings the sessions of all your coding assistants onto one screen. It reads what the harnesses already write to disk and turns it into sessions, tokens, costs and activity you can reason about.
The CLI is called agentop: a single binary that runs the local server, the web interface,
the terminal TUI and the MCP server.
Harnesses
Section titled “Harnesses”agentistics is plural: it tracks several harnesses side by side, without depending on any of them.
- Claude Code
- Codex
- Gemini
- Copilot
- Antigravity
- Kimi
- OpenCode
When a harness cannot produce a metric (for example, Gemini’s local files carry no token data), it
shows as N/A, never as a 0 that looks true.
What you get
Section titled “What you get”- The session fleet. What is running now, what is waiting on you and what can be reopened, from every harness, in one place.
- Costs with an owner. Cost and usage per session, project, repository, harness and model.
- Agentask. The task board: each task gathers the sessions that worked on it and shows what the delivery cost, in how many rounds and across how many sessions.
- Several front doors onto the same fleet. The web interface, the terminal cockpit, the
agentop sessioncommand, the VS Code extension and the MCP for your own agents. - Export. Metrics via OpenTelemetry and PDF reports.
- PT-BR and EN, light and dark themes, in the CLI, the TUI and the web.
Where it runs
Section titled “Where it runs”| System | How to install |
|---|---|
| Linux (x86_64) and WSL | curl script or npm |
| macOS (x86_64 and arm64) | curl script |
| Windows | .msi or .exe installer |
The session manager uses tmux, so on Windows it runs inside WSL.
Next steps
Section titled “Next steps”The code is at github.com/agentistics/agentistics.