README.md

OpenModelStudio plugin for OpenClaw

This OpenClaw plugin registers agent tools that call the OpenModelStudio REST API. AI agents can create projects, models, start training, manage experiments, launch workspaces, and search—so users can say “create a project and train a model” in Telegram/WhatsApp/Discord and the agent does it.

Prerequisites

• OpenClaw installed and configured (Node 22+).
• OpenModelStudio API running (e.g. make k8s-deploy then API at http://localhost:31001).
• An API key (recommended) or JWT from OpenModelStudio (see below).

Getting an access token

Recommended — API key (no expiry; end users never provide a token):

1. Log into the OpenModelStudio UI.
2. Go to Settings → API Keys and create a key (e.g. name: "OpenClaw").
3. Copy the key (it looks like oms_xxxxxxxx...; you only see it once).
4. Use it as accessToken in the plugin config. No refresh needed.

Alternatively — JWT (short-lived):

curl -s -X POST http://localhost:31001/auth/login \
  -H "Content-Type: application/json" \
  -d '{"email":"test@openmodel.studio","password":"Test1234"}' | jq -r '.access_token'

Use the printed token as accessToken. JWTs expire (e.g. 15–60 min); refresh when you get 401s.

Installation

From the repo (local path)

# From OpenModelStudio repo root
openclaw plugins install -l ./integrations/openclaw

Or copy into OpenClaw’s extensions dir:

cp -r integrations/openclaw ~/.openclaw/extensions/openmodelstudio

From npm (if published)

openclaw plugins install @openmodelstudio/openclaw-plugin

Install plugin dependencies:

cd ~/.openclaw/extensions/openmodelstudio  # or your install path
npm install

Configuration

Add to your OpenClaw config (e.g. ~/.openclaw/openclaw.json or Control UI):

{
  plugins: {
    entries: {
      "openclaw-plugin": {
        enabled: true,
        config: {
          baseUrl: "http://localhost:31001",
          accessToken: "YOUR_JWT_OR_TOKEN"
        }
      }
    }
  }
}

(The entry key is openclaw-plugin, from the package name.)

• baseUrl — OpenModelStudio API base (no trailing slash). Examples: http://localhost:31001, https://api.openmodel.studio.
• accessToken — API key (e.g. oms_...) or JWT; stored as sensitive in the UI. Prefer an API key so users don’t have to provide or refresh a token.

Production: Use HTTPS for the API URL and practice good secret handling (secrets manager or secure config, restrict permissions on openclaw.json, never commit credentials).

Because the tools are optional, you must allowlist them for your agent. For example:

{
  agents: {
    list: [
      {
        id: "main",
        tools: {
          allow: [
            "oms_list_projects",
            "oms_create_project",
            "oms_list_models",
            "oms_create_model",
            "oms_start_training",
            "oms_list_training_jobs",
            "oms_get_training_job",
            "oms_get_job_metrics",
            "oms_create_experiment",
            "oms_add_experiment_run",
            "oms_list_experiment_runs",
            "oms_list_experiments",
            "oms_launch_workspace",
            "oms_list_datasets",
            "oms_search"
          ]
        }
      }
    ]
  }
}

Or allow all tools from this plugin by id:

agents: {
  list: [{ id: "main", tools: { allow: ["openclaw-plugin"] } }]
}

Restart the OpenClaw Gateway after changing config.

Tools

Tool	Description
oms_list_projects	List projects the user can access
oms_create_project	Create a project (name, optional description)
oms_list_models	List models (optional project_id)
oms_create_model	Create a model (project_id, name, framework, optional description/source_code)
oms_start_training	Start training (model_id; optional dataset_id, hardware_tier, hyperparameters JSON)
oms_list_training_jobs	List recent training jobs
oms_get_training_job	Get job status by job_id
oms_get_job_metrics	Get stored metrics for a job
oms_create_experiment	Create experiment (project_id, name, optional description)
oms_add_experiment_run	Add run to experiment (experiment_id, job_id; optional parameters/metrics JSON)
oms_list_experiment_runs	List runs for an experiment
oms_list_experiments	List experiments (optional project_id)
oms_launch_workspace	Launch JupyterLab workspace (project_id; optional name, hardware_tier)
oms_create_dataset	Create a dataset (project_id, name, format e.g. csv; optional description)
oms_list_datasets	List datasets (optional project_id)
oms_search	Search projects, models, datasets by query

All IDs are UUIDs returned by the API (e.g. from oms_create_project, oms_create_model, oms_start_training).

Agent instructions (SOUL.md / AGENTS.md)

So the agent uses the tools in the right order and chains steps (create project → create model → start training, etc.), add workflow instructions to your OpenClaw workspace. A full example is in AGENTS-oms-example.md in this plugin folder—copy or paste its content into your workspace AGENTS.md (or SOUL.md). Minimal version:

You have access to OpenModelStudio (OMS) tools. Use them to manage ML projects:

1. Create a project with oms_create_project, then use the returned project id.
2. Create a model with oms_create_model (project_id, name, framework).
3. Start training with oms_start_training (model_id; optionally dataset_id, hyperparameters).
4. Create an experiment with oms_create_experiment, then add runs with oms_add_experiment_run (experiment_id, job_id, parameters, metrics).
5. Launch a workspace with oms_launch_workspace (project_id) for interactive JupyterLab.

Always use the exact ids returned by the API (project_id, model_id, job_id, experiment_id) in later steps. After each action, confirm what you did and tell the user they can check the OpenModelStudio dashboard.

User flow

1. User talks to the OpenClaw agent (e.g. in Telegram): “Create a project called My ML and start a training job for the default model.”
2. Agent calls oms_list_projects, then oms_create_project, then oms_list_models, then oms_start_training (using returned ids).
3. Agent replies: “Created project My ML and started training. You can watch the run in OpenModelStudio at http://localhost:31000.”
4. User sees the same project and job in the OpenModelStudio UI.

License

Same as OpenModelStudio (GPL-3.0).