Cocktail (c8ctl) - Camunda 8 CLI

c8ctl (pronounced: "cocktail") — a minimal-dependency CLI for Camunda 8 operations built on top of @camunda8/orchestration-cluster-api.

Features

• Multi-Tenant Support: Full support for multi-tenancy across all operations
• Profile Management: Store and manage multiple cluster configurations
• Camunda Modeler Integration: Automatically import and use profiles from Camunda Modeler
• Identity Management: Manage users, roles, groups, tenants, authorizations, and mapping rules
• Plugin System: Extend c8ctl with custom commands via npm packages
• Local Cluster: Run a local Camunda 8 instance directly — no Docker required
• MCP Proxy: Bridge local MCP clients to remote Camunda 8 for AI agent integration
• Building Block Deployment: Automatic prioritization of *_bb-* folders during deployment, marked with 🧱 in results
• Process Application Support: Resources in folders with .process-application file marked with 📦 in results
• Enhanced Deployment Results: Table view showing file paths, visual indicators, resource details, and versions
• Watch Mode: Monitors a folder for file changes and auto-redeploys (configurable extensions via --extensions)
• .c8ignore Support: Filter deploy/watch file scanning with .gitignore-style patterns; node_modules/, target/, .git/ ignored by default
• Open Applications: Open Camunda web applications (Operate, Tasklist, Modeler, Optimize) in the browser directly from the CLI
• Search: Powerful search across process definitions, process instances, user tasks, incidents, jobs, and variables with filter, wildcard, and case-insensitive support
• Flexible Output: Switch between human-readable text and JSON output modes
• Shell Completion: Auto-install completions for bash, zsh, and fish with automatic refresh on upgrade
• Agent Flags: --dry-run to preview API requests and --fields to filter output columns for AI agents and scripts
• Debug Mode: Enable detailed logging to stderr with DEBUG=1 or C8CTL_DEBUG=true

Beware the 🤖

Full transparency:
this cli is also a pilot-coding experiment, practicing Agentic Engineering.
Guided by humans, the codebase is (mostly) built by your friendly neighborhood LLM, fully dogfooding the Human-in-the-Loop pattern.

Installation

Requirements

• Node.js >= 22.18.0 (for native TypeScript support)

Global Installation (Recommended)

npm install @camunda8/cli -g

After installation, the CLI is available as c8ctl (or its alias c8).

Note: The c8 alias provides typing ergonomics for common keyboard layouts - the c key (left index finger) followed by 8 (right middle finger) makes for a comfortable typing experience on both QWERTY and QWERTZ keyboards.

Documentation

The full documentation is maintained in this repository under docs/ and is published — and kept in sync — on docs.camunda.io:

• Getting started — install, spin up a local cluster, credentials, profiles, output modes, and shell completion.
• Cluster inspection and process management — list, search, get, and manage process instances, user tasks, incidents, jobs, variables, messages, and forms.
• Identity management — users, roles, groups, tenants, authorizations, and mapping rules.
• Development workflows — deploy, run, watch, profiles, sessions, and the MCP proxy.
• Plugins — scaffold, install, and manage custom commands.
• Command reference — every command, flag, resource, and alias.

The docs/ pages are the single source of truth. When you change CLI behavior, update the relevant page here; a workflow syncs them to camunda-docs.

Quick start

# Install
npm install @camunda8/cli -g

# Start a local Camunda 8 cluster (no Docker required)
c8ctl cluster start

# Deploy a process and start an instance
c8ctl deploy ./my-process.bpmn
c8ctl run my-process

# Inspect running instances
c8ctl list process-instances

# Show help for any command
c8ctl help
c8ctl help <command>

For everything else, see the documentation above.

AI agents and scripting

c8ctl ships flags built for AI agents and scripts — --fields (filter output columns), --dry-run (preview the API request without executing it), and machine-readable JSON help. They appear in their own section in c8ctl help.

See Development workflows → AI agents and scripting for details and examples. For a full machine-readable reference intended for agents, see CONTEXT.md.

Command Structure

c8ctl <verb> <resource> [arguments] [flags]

Verbs:

• list - List resources (process, identity)
• search - Search resources with filters
• get - Get resource by key
• create - Create resource
• delete - Delete resource
• cancel - Cancel resource
• await - Create and await completion (alias for create --awaitCompletion)
• complete - Complete resource
• fail - Fail a job
• update - Update a job
• activate - Activate jobs by type
• resolve - Resolve incident
• publish - Publish message
• correlate - Correlate message
• set - Set variables on an element instance
• deploy - Deploy resources
• run - Deploy and start process
• assign - Assign resource to target
• unassign - Unassign resource from target
• watch (alias: w) - Watch files for changes and auto-deploy
• open - Open Camunda web application in browser
• add - Add a profile
• remove (alias: rm) - Remove a profile
• load - Load a c8ctl plugin
• unload (alias: rm) - Unload a c8ctl plugin
• upgrade - Upgrade a plugin
• downgrade - Downgrade a plugin to a specific version
• sync - Synchronize plugins
• init - Create a new plugin from TypeScript template
• doctor - Diagnose plugin loading state and collisions
• use - Set active profile or tenant
• output - Show or set output format
• completion - Generate shell completion script
• mcp-proxy - Start a STDIO to remote HTTP MCP proxy server
• feedback - Open the feedback page to report issues or request features
• help (alias: menu) - Show help
• which - Show active profile or output mode

Resources: authorization (auth), form, group, incident (inc), job, jobs, mapping-rule (mr), message (msg), plugin, process-definition (pd), process-instance (pi), profile, role, tenant, topology, user, user-task (ut), variable (var, vars), wait-state (ws)

Tip: Run c8ctl help <command> to see detailed help for specific commands with all available flags.

Environment Variables

• CAMUNDA_BASE_URL: Cluster base URL
• CAMUNDA_CLIENT_ID: OAuth client ID
• CAMUNDA_CLIENT_SECRET: OAuth client secret
• CAMUNDA_TOKEN_AUDIENCE: OAuth token audience
• CAMUNDA_OAUTH_URL: OAuth token endpoint
• CAMUNDA_DEFAULT_TENANT_ID: Default tenant ID
• C8CTL_OUTPUT_MODE: Per-invocation output mode override (json or text); does not persist. Lower precedence than --json.
• C8CTL_DATA_DIR: Override the OS-default data directory for plugins and session state.
• C8CTL_DEBUG / DEBUG: Enable debug logging to stderr.

Configuration Files

c8ctl Configuration

Configuration is stored in platform-specific user data directories:

• Linux: ~/.config/c8ctl/
• macOS: ~/Library/Application Support/c8ctl/
• Windows: %APPDATA%\c8ctl\

Files:

• profiles.json: Saved cluster configurations
• session.json: Active profile, tenant, and output mode
• plugins.json: Plugin registry tracking installed plugins

Camunda Modeler Configuration

c8ctl automatically reads profiles from Camunda Modeler (if installed):

• Linux: ~/.config/camunda-modeler/profiles.json
• macOS: ~/Library/Application Support/camunda-modeler/profiles.json
• Windows: %APPDATA%\camunda-modeler\profiles.json

Modeler profiles are:

• Read-only in c8ctl (managed via Camunda Modeler)
• Automatically loaded on each command execution
• Prefixed with modeler: when used in c8ctl
• Support both cloud and self-managed clusters

Command Reference

A compact list of every verb and resource is in Command Structure above.

For the complete reference of every command, flag, resource, and alias, see:

• docs/command-reference.md in this repository, or
• the c8ctl command reference on docs.camunda.io.

License

Apache 2.0 - see LICENSE.md

Contributing

See CONTRIBUTING.md for development setup, testing, project structure, and how to add new commands. See AGENTS.md for commit conventions and coding standards.