Lumen CLI Manual In development
lumen-cli is Lumen’s optional CLI (built in Rust), providing command-line workflow capabilities that don’t depend on the Python SDK.
Installation
bash
cargo install lumen-cli
# 或
curl -fsSL https://lumen.lurus.cn/install.sh | shOverview
lumen <command> [options]| Command | Purpose |
|---|---|
doctor | Environment self-check: token, network, disk, Python dependencies |
init | Generate a lumen.yaml config template in the project root |
agent | Agent-level operations: list / trace / replay / export |
mcp | MCP compatibility layer: expose Lumen capabilities to Claude/Codex |
workflow | Workflow executor based on lumen.yaml |
deploy | Push Agent definitions to the Kova Cluster |
config | View/modify local CLI configuration |
Command Reference
bash
# doctor — 环境自检（退出码 0 全通过 / 1 至少一项失败）
lumen doctor
# ✓ LURUS_API_KEY present ✓ python3.11 detected ✓ /var/lumen writable (5.2 GB free)
# ✗ port 7070 occupied — close the process or set LUMEN_PORT
# init — 项目根生成 lumen.yaml（templates: langgraph / bare / multi-agent）
lumen init --template langgraph
# agent
lumen agent list # 列出本地/远端 Agent
lumen agent trace <run-id> # 打印 trace 树
lumen agent replay <run-id> # 不消耗 Token 重放
lumen agent export <run-id> # 导出 JSON / HAR / OTel
# mcp — 启动 MCP 服务端，暴露 Trace/Replay/Cost 为工具
lumen mcp serve --port 3333 # 或 --manifest ./my-tools.yaml
# Claude Code / Codex 的 mcp_servers 指向 http://127.0.0.1:3333 即可调用
# workflow — 按 lumen.yaml 执行多 Agent 编排
lumen workflow run # 默认 pipeline；-e prod 指定环境；--dry-run 不实际调 LLM
# deploy — 推送 Agent 定义到 Kova Cluster
lumen deploy --target kova://my-cluster
# config
lumen config get api_key
lumen config set api_key sk-xxxx
lumen config unset telemetry.endpointBoth replay and dry-run are free
lumen agent replay <run-id> replays from history records and does not consume tokens; lumen workflow run --dry-run walks through the full orchestration flow but does not actually call the LLM, making it ideal for validating your lumen.yaml configuration.Exit Codes
| Code | Meaning |
|---|---|
0 | Success |
1 | Generic error |
2 | Argument error |
3 | Missing configuration |
4 | Network error |
5 | Remote service returned an error |