LoongSuite Pilot CLI
A local telemetry collector for AI coding agents: tokens, sessions and tool calls in one OpenTelemetry-style event schema
- 240 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Installation writes hooks and plugins into your agents' configuration, and they see prompts and tool arguments
- With export on, prompts and replies go to SLS, an HTTP endpoint or an OTLP backend, and they can hold secrets and personal data
- The installer is downloaded by a script from an Alibaba Cloud OSS bucket and run with `bash`
Install
- Works in
Manual install
curl -fsSL https://loongcollector-community-edition.oss-cn-shanghai.aliyuncs.com/loongsuite-pilot/installer.sh -o /tmp/loongsuite-pilot-installer.sh && bash /tmp/loongsuite-pilot-installer.sh installFor Linux and macOS. On Windows use `installer.ps1` from the same location. After installing run `loongsuite-pilot status`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Installation writes hooks and plugins into your agents' configuration, and they see prompts and tool arguments
- With export on, prompts and replies go to SLS, an HTTP endpoint or an OTLP backend, and they can hold secrets and personal data
- The installer is downloaded by a script from an Alibaba Cloud OSS bucket and run with `bash`
This is third-party code. Review the repository files before installing.
What it does
LoongSuite Pilot discovers supported AI agents on a developer machine, installs hooks or plugins in them and collects their activity into a shared GenAI event schema. The data goes out as logs or traces: local JSONL, Alibaba Cloud SLS, an arbitrary HTTP endpoint or an OTLP backend. The list of supported agents is long: Claude Code, Codex, Cursor, GitHub Copilot CLI, OpenCode, Hermes Agent, OpenClaw, Qwen Code, Qoder, Kiro CLI and others. Agent definitions live in the `agents.d/` folder, so a new agent is added without touching the core. The `loongsuite-pilot` command manages the service: `start`, `stop`, `restart`, `status`, `info`, `token-usage` and `rollback`. A local dashboard at `http://127.0.0.1:8765/` starts with the service and shows tokens, sessions, requests, tools, models and repository activity. A content capture policy is set per agent, and secrets are masked before export. On macOS a menu bar app shows live counters.
Who it is for. Development teams and platform engineers running several AI agents who need token spend, sessions and tool calls in one place.
Good fit when
- A team uses Claude Code, Cursor and Codex at once and has to pull token spend out of three log formats
- A manager asks for a report on which agents, models and tools actually work on which repositories
- Agent events have to be exported to Alibaba Cloud SLS or an OTLP-compatible tracing system
- Secrets must be masked before export and the amount of prompt content leaving the machine limited
- Collection for one more agent has to be added: the definition goes into `agents.d/` per the Agent Onboarding guide
Not a fit when
- You only need token and cost counting from local logs with no hooks: ccusage does that
- Telemetry has to go into your own application: Pilot collects agent data on the machine and does not instrument your code, and the catalog has OpenTelemetry Agent Skills for code
- The team does not want a tool changing agent settings: the installer writes hooks and plugins into their configuration
Example request
Show the loongsuite-pilot status and how many tokens Claude Code and Codex used todayLimitations
Node.js that can load `node:sqlite` is required: versions 22.13, 23.4 and newer; the installer downloads a managed Node.js 22.22. Automatic installation for GitHub Copilot CLI works only in Unix shells, and Windows and subagents are not supported. On Windows Claude Code, Cursor, Qoder, OpenCode and WorkBuddy are explicitly documented, and for the other agents support is not stated. Token usage is not collected for Kiro CLI and Doubao Work, and Doubao Work stays experimental and returns metadata only.
How to disable. Stop the service with `loongsuite-pilot stop`. Uninstalling Pilot and its hooks is covered in `docs/installation.md`. Local logs live in `~/.loongsuite-pilot/logs/output/` and are deleted by hand.
FAQ
Where does the data go by default?
Only to local JSONL files in `~/.loongsuite-pilot/logs/output/` on Linux and macOS and in `%USERPROFILE%\.loongsuite-pilot\logs\output\` on Windows. SLS, HTTP and OTLP are connected separately in the configuration.
Does Pilot store prompt text?
That is set by a per-agent content capture policy, and secrets are masked before output. Content fields in the event schema are marked sensitive and are opt-in.
How do I open the dashboard?
It starts with the service and is available at `http://127.0.0.1:8765/`. There is no separate command for it. On macOS `loongsuite-pilot dashboard shortcut install` creates a Dock shortcut.
Can events be linked to an external trace?
Yes, linking to a parent trace is turned on with the `upstreamLink.enabled` setting or the `LOONGSUITE_PILOT_UPSTREAM_LINK` variable. It is off by default and does not affect normal collection.
In what order is configuration read?
Environment variables first, then `~/.loongsuite-pilot/config.json`, then built-in defaults.
Discussion
No questions or reviews yet.
Related
CLI toolsOpenTelemetry-based CLI tracing for Claude Code, Cursor and Codex sessions, showing what the agent actually did and what it cost
A local dashboard for Claude Code, Codex CLI, Gemini CLI, Copilot and OpenCode that shows tokens, cost, latency and errors
A self-hosted control center for coding agents: Claude Code, Codex, Copilot, OpenCode and Pi from your phone, desktop and terminal
A local proxy between AI agents and model providers that prices every run as it happens and cuts off spend above your cap
Tokscale
tokscale
A CLI with an interactive TUI that tracks token usage and cost across AI coding agents: Claude Code, Codex, Copilot, Cursor and more
Agent of Empires
Agent of Empires (AoE)
A session manager for coding agents: runs Claude Code, OpenCode, Codex and other CLIs in parallel, isolated tmux sessions with a TUI and web dashboard
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| LoongSuite Pilot (this card) | yes | 240 | |
| ccusage | no | 19,337 | counts tokens and cost from local logs with no hooks, while Pilot installs hooks and exports events to JSONL, SLS, HTTP and OTLP |
| OpenTelemetry Agent Skills by OllyGarden | no | 125 | teaches the agent to wire telemetry into your application, while Pilot collects telemetry from the agents themselves |
README in short
The README explains the problem: each agent records activity in its own local format, and Pilot reduces them to one schema. It continues with a table of supported agents, prerequisites, installation for Linux, macOS and Windows, output and privacy setup and a command reference.