Preface¶
In agent development environments like DSH, where “everything is a plugin,” model calls, training tasks, evaluations, approvals, traces, and usage information are often scattered across different tools. The DSH community directory is an independent site and has no official affiliation with DeepSeek / High-Flyer.
Below, we introduce cerebrixos-org/tuning-engines-cli. It provides a CLI, MCP server, and Python agent runtime adapters for managing models, agents, skills, and MCP workflows.
What This Is¶
cerebrixos-org/tuning-engines-cli is maintained by cerebrixos-org and is licensed under MIT. The package.json reports version 0.4.35 and requires node >=18.0.0.
One-line positioning: It brings inference routing, permission policies, approvals, traces, usage, and model training management into a single CLI / MCP entry point.
Core Capabilities¶
Below are several verified capabilities:
- A CLI and built-in MCP server providing 60+ tools.
- Routes inference through a single OpenAI-compatible API with support for RBAC, traffic policies, approvals, traces, and usage.
- Fine-tunes open-source models on domain data using agents such as Cody (
code_repo) and SIERA (sera_code_repo). - Manages training jobs, evaluations, inference usage, traces, approvals, and non-secret tenant registry metadata.
- Initializes orchestration starters for LangGraph, Temporal, inngest, triggerdev, hatchet, restate, dbos, dapr, prefect, dagster, and airflow.
- Provides Work Sessions / outcomes, along with optional native telemetry hooks for Claude Code or Codex.
- Offers a Claude Code plugin wrapper around the same MCP server.
Installation and Enablement¶
First, confirm that the Node.js version meets node >=18.0.0, then install the CLI:
npm install -g tuningengines-cli
After installation, log in to your account:
te auth login
The verified command to start the MCP server is:
npx -y --package tuningengines-cli@latest te mcp serve
The verified command to add the MCP service in Claude Code is:
claude mcp add tuning-engines -- npx -y --package tuningengines-cli@latest te mcp serve
The repository also provides a Claude Code plugin wrapper. The verified command includes:
claude plugin marketplace add cerebrixos-org/tuning-engines-cli
For the complete Claude Code plugin installation sequence, refer to the repository README; this article does not assemble unprovided installation commands.
Before enabling this in DSH, you should inspect the source code and license. Once enabled, the plugin will run with the permissions of the current dsh process; you should verify the resource scope accessible to that process before enabling. The MCP configuration preserves the TE_API_KEY environment variable boundary.
Typical Usage¶
Below are verified command examples.
Create a training job using the code_repo agent:
te jobs create --agent code_repo --base-model Qwen/Qwen2.5-Coder-7B-Instruct --repo-url https://github.com/your-org/your-repo --output-name my-model
Track job status:
te jobs status <job-id> --watch
Initialize a LangGraph orchestration starter:
te orchestration init langgraph
Start a Work Session:
te goal start "Fix flaky checkout retries"
Usage Boundaries and Notes¶
The MCP server intentionally does not expose internal proxy routes and rejects MCP-side inference-key creation and raw secret-bearing mutation fields.
If you need to use one-time keys, raw provider secrets, S3 credential validation, or import/export S3 assets with raw credentials, you should use the CLI or web UI.
If you use Claude Code hooks, project-local hooks will write to .claude/settings.local.json, which may require a restart; Codex project hooks need to be reviewed and trusted from /hooks.
Conclusion¶
The value of cerebrixos-org/tuning-engines-cli lies in bringing model fine-tuning, inference governance, orchestration initialization, and Claude Code / Codex hooks into a single CLI/MCP entry point. It suits developers who need to centrally manage multiple model workflows within the DSH plugin ecosystem.
Directory page:
https://www.skillhub.cn/plugins/cerebrixos-org/tuning-engines-cli
GitHub:
https://github.com/cerebrixos-org/tuning-engines-cli
