CLI

The CableKnit CLI lets you build, validate, push, and monitor plugins from your terminal.

The easiest way to get started is to run the cableknit scaffold generator in an empty folder, next start the MCP server, then run your AI agent in that same folder and have a conversation with your AI about building a CableKnit app in that folder.

## Install the CLI

### Homebrew (macOS and Linux)

```bash
brew install jessewaites/cableknit/cableknit
```

### Direct download

Download the latest binary from the [GitHub Releases](https://github.com/jessewaites/cableknit-cli/releases) page. Builds are available for macOS (Apple silicon), Linux, and Windows.

### Go install

```bash
go install github.com/jessewaites/cableknit-cli@latest
```

Verify the installation:

```bash
cableknit version
```

## Authenticate

Before using most commands you need to log in with your CableKnit account.

### Log in

```bash
cableknit login
```

You see an interactive form that prompts you for your email and password. On success you see your account details in a confirmation box.

### Check your session

```bash
cableknit whoami
```

Prints your name, email, and publisher status.

### Log out

```bash
cableknit logout
```

Clears your local session and invalidates the token on the server.

## Develop plugins

### Generate a new plugin

Scaffold a new plugin project with all the standard folders and starter files:

```bash
cableknit generate
```

Aliases: `init`, `scaffold`

An interactive form prompts you for: plugin name, description, category, industry, and billing interval. The CLI creates a ready-to-edit bundle:

```
<slug>/
  plugin.json
  skills/
  automations/
  blueprints/
  tools/
    sample-lookup.json     # data_store example
    reference-data.json    # static data example
  docs/
    getting-started.md
  images/
```

### Validate a bundle

Run validation against the CableKnit platform without publishing anything:

```bash
# Validate the current directory (must contain plugin.json)
cableknit validate

# Validate a specific path
cableknit validate ./my-plugin

# Validate a zip file
cableknit validate my-plugin.zip
```

The CLI will zip the directory (if needed), upload it, and report any errors or warnings. A successful validation prints a summary of your plugin's name, slug, version, skills, automations, and docs.

### Push a bundle

Publish or update your plugin:

```bash
cableknit push

cableknit push ./my-plugin
```

The CLI tells you whether a new plugin was created or an existing one was updated.

## Monitor runs

### List plugins

```bash
cableknit plugins list
```

Displays an interactive table with columns for name, slug, version, status, visibility, and install count.

### List runs

```bash
# All runs
cableknit runs list

# Filter by status
cableknit runs list --status running

# Limit results
cableknit runs list --limit 10
```

Statuses are color-coded: running (blue), paused (yellow), completed (green), failed (red), and cancelled (dim).

### Tail a run

Stream live logs from a running automation:

```bash
cableknit runs tail RUN-7F3A
```

This opens a full-screen view with a scrolling log, an elapsed-time counter, and a live spinner. The view freezes and shows a summary when the run reaches a terminal state. Press `ctrl+c` to exit early.

For plain-text output (useful for piping or CI):

```bash
cableknit runs tail RUN-7F3A --no-tui
```

## Discover connectors

### List available connectors

```bash
cableknit connectors list
```

Displays a table with columns: NAME, SLUG, CATEGORY, AUTH, STATUS. Supports `--json` for scripted output.

Use this to discover connector slugs when setting `requirements.connectors` in your `plugin.json`.

## Observability

View logs, metrics, and errors for your published plugins. See the full [Monitoring guide](/monitoring) for details.

```bash
# View recent logs
cableknit logs <plugin-slug>

# View metrics dashboard
cableknit metrics <plugin-slug>

# View errors only
cableknit errors <plugin-slug>
```

All three support `--json` for scripted output.

## Versioning

View version history and roll back to previous bundle versions. See the full [Versioning guide](/versioning) for details.

```bash
# View version history
cableknit versions <plugin-slug>

# Roll back to a specific version
cableknit rollback <plugin-slug> --version 1.1.0

# Interactive rollback (pick from list)
cableknit rollback <plugin-slug>
```

Both commands support `--json` for scripted output.

## Global flags

These flags work with any command:

| Flag | Description |
|------|-------------|
| `--api-url` | Override the API base URL |
| `--no-color` | Disable color output |
| `--debug` | Enable debug logging |
| `--json` | Output structured JSON instead of styled text |

### JSON mode

Any data command supports `--json` for scripting and automation:

```bash
cableknit plugins list --json
cableknit runs list --json
cableknit validate --json
```

## AI editor integration (MCP)

The CLI includes a built-in [MCP](https://modelcontextprotocol.io) server that gives AI-powered editors deep knowledge of the CableKnit platform. Your editor's AI assistant can look up schemas, generate valid plugin code, validate bundles, explain errors, describe automation flows, and walk you through building a complete plugin.

Works with Claude Code, Cursor, Windsurf, and any MCP-compatible editor.

### Auto-install

```bash
cableknit mcp install
```

This detects your installed editors and configures them automatically. Or target a specific editor:

```bash
cableknit mcp install claude
cableknit mcp install cursor
cableknit mcp install windsurf
```

Restart your editor and the AI assistant will have access to all CableKnit tools.

### Manual setup

<Tabs>
  <Tab title="Claude Code">
    Run this one-liner:
    ```bash
    claude mcp add cableknit -- cableknit mcp
    ```
  </Tab>
  <Tab title="Cursor">
    Add to `.cursor/mcp.json`:
    ```json
    {
      "mcpServers": {
        "cableknit": {
          "command": "cableknit",
          "args": ["mcp"]
        }
      }
    }
    ```
  </Tab>
  <Tab title="Windsurf">
    Add to `.windsurf/mcp.json`:
    ```json
    {
      "mcpServers": {
        "cableknit": {
          "command": "cableknit",
          "args": ["mcp"]
        }
      }
    }
    ```
  </Tab>
</Tabs>

### What your AI assistant gets

| Category | Count | Examples |
|----------|-------|---------|
| Reference tools | 15 | Schema lookups, platform tools, connectors, industries, categories, bundle limits, docs |
| Validation tools | 5 | Validate plugin.json, validate automations, explain errors, check prompt injection, describe automation flows |
| Generation tools | 7 | Generate plugin.json, automations, blueprints, data sources, skills, doc pages, scaffold full bundles |
| Resources | 8 | Schemas, CLI guide, app guidelines, sample plugin |
| Prompt templates | 4 | Build a plugin, debug validation, add automation, review before submit |

No authentication required — all tools work without logging in.

### Example usage

After installing, ask your editor's AI assistant:

- *"What's the plugin.json schema for CableKnit?"*
- *"Generate a CableKnit plugin that routes support tickets for healthcare"*
- *"Explain this validation error: slug invalid format"*
- *"Describe what this automation does"* (paste your workflow JSON)
- *"Review my plugin before I submit"*

## Configuration

The CLI stores its config at `~/.cableknit/config.yaml`. This file holds your auth token and any overrides. You generally don't need to edit it directly — `cableknit login` and `cableknit logout` manage it for you.

The CableKnit CLI lets you build, validate, push, and monitor plugins from your terminal.

Install the CLI

Homebrew (macOS and Linux)

brew install jessewaites/cableknit/cableknit

Direct download

Download the latest binary from the GitHub Releases page. Builds are available for macOS (Apple silicon), Linux, and Windows.

Go install

go install github.com/jessewaites/cableknit-cli@latest

Verify the installation:

cableknit version

Authenticate

Before using most commands you need to log in with your CableKnit account.

Log in

cableknit login

You see an interactive form that prompts you for your email and password. On success you see your account details in a confirmation box.

Check your session

cableknit whoami

Prints your name, email, and publisher status.

Log out

cableknit logout

Clears your local session and invalidates the token on the server.

Develop plugins

Generate a new plugin

Scaffold a new plugin project with all the standard folders and starter files:

cableknit generate

Aliases: init, scaffold

An interactive form prompts you for: plugin name, description, category, industry, and billing interval. The CLI creates a ready-to-edit bundle:

<slug>/
  plugin.json
  skills/
  automations/
  blueprints/
  tools/
    sample-lookup.json     # data_store example
    reference-data.json    # static data example
  docs/
    getting-started.md
  images/

Validate a bundle

Run validation against the CableKnit platform without publishing anything:

# Validate the current directory (must contain plugin.json)
cableknit validate

# Validate a specific path
cableknit validate ./my-plugin

# Validate a zip file
cableknit validate my-plugin.zip

The CLI will zip the directory (if needed), upload it, and report any errors or warnings. A successful validation prints a summary of your plugin’s name, slug, version, skills, automations, and docs.

Push a bundle

Publish or update your plugin:

cableknit push

cableknit push ./my-plugin

The CLI tells you whether a new plugin was created or an existing one was updated.

Monitor runs

List plugins

cableknit plugins list

Displays an interactive table with columns for name, slug, version, status, visibility, and install count.

List runs

# All runs
cableknit runs list

# Filter by status
cableknit runs list --status running

# Limit results
cableknit runs list --limit 10

Statuses are color-coded: running (blue), paused (yellow), completed (green), failed (red), and cancelled (dim).

Tail a run

Stream live logs from a running automation:

cableknit runs tail RUN-7F3A

This opens a full-screen view with a scrolling log, an elapsed-time counter, and a live spinner. The view freezes and shows a summary when the run reaches a terminal state. Press ctrl+c to exit early.

For plain-text output (useful for piping or CI):

cableknit runs tail RUN-7F3A --no-tui

Discover connectors

List available connectors

cableknit connectors list

Displays a table with columns: NAME, SLUG, CATEGORY, AUTH, STATUS. Supports --json for scripted output.

Use this to discover connector slugs when setting requirements.connectors in your plugin.json.

Observability

View logs, metrics, and errors for your published plugins. See the full Monitoring guide for details.

# View recent logs
cableknit logs <plugin-slug>

# View metrics dashboard
cableknit metrics <plugin-slug>

# View errors only
cableknit errors <plugin-slug>

All three support --json for scripted output.

Versioning

View version history and roll back to previous bundle versions. See the full Versioning guide for details.

# View version history
cableknit versions <plugin-slug>

# Roll back to a specific version
cableknit rollback <plugin-slug> --version 1.1.0

# Interactive rollback (pick from list)
cableknit rollback <plugin-slug>

Both commands support --json for scripted output.

Global flags

These flags work with any command:

Flag	Description
`--api-url`	Override the API base URL
`--no-color`	Disable color output
`--debug`	Enable debug logging
`--json`	Output structured JSON instead of styled text

JSON mode

Any data command supports --json for scripting and automation:

cableknit plugins list --json
cableknit runs list --json
cableknit validate --json

AI editor integration (MCP)

The CLI includes a built-in MCP server that gives AI-powered editors deep knowledge of the CableKnit platform. Your editor’s AI assistant can look up schemas, generate valid plugin code, validate bundles, explain errors, describe automation flows, and walk you through building a complete plugin.

Works with Claude Code, Cursor, Windsurf, and any MCP-compatible editor.

Auto-install

cableknit mcp install

This detects your installed editors and configures them automatically. Or target a specific editor:

cableknit mcp install claude
cableknit mcp install cursor
cableknit mcp install windsurf

Restart your editor and the AI assistant will have access to all CableKnit tools.

Manual setup

Run this one-liner: ```bash claude mcp add cableknit -- cableknit mcp ``` Add to `.cursor/mcp.json`: ```json { "mcpServers": { "cableknit": { "command": "cableknit", "args": ["mcp"] } } } ``` Add to `.windsurf/mcp.json`: ```json { "mcpServers": { "cableknit": { "command": "cableknit", "args": ["mcp"] } } } ```

What your AI assistant gets

Category	Count	Examples
Reference tools	15	Schema lookups, platform tools, connectors, industries, categories, bundle limits, docs
Validation tools	5	Validate plugin.json, validate automations, explain errors, check prompt injection, describe automation flows
Generation tools	7	Generate plugin.json, automations, blueprints, data sources, skills, doc pages, scaffold full bundles
Resources	8	Schemas, CLI guide, app guidelines, sample plugin
Prompt templates	4	Build a plugin, debug validation, add automation, review before submit

No authentication required — all tools work without logging in.

Example usage

After installing, ask your editor’s AI assistant:

“What’s the plugin.json schema for CableKnit?”
“Generate a CableKnit plugin that routes support tickets for healthcare”
“Explain this validation error: slug invalid format”
“Describe what this automation does” (paste your workflow JSON)
“Review my plugin before I submit”

Configuration

The CLI stores its config at ~/.cableknit/config.yaml. This file holds your auth token and any overrides. You generally don’t need to edit it directly — cableknit login and cableknit logout manage it for you.