MCP Server

fibe.gg includes a built-in Model Context Protocol (MCP) server. Any MCP-compatible AI agent or client can connect to the MCP server and manage the entire platform using natural language.

Quick Start

1. Create an API Key

Create an API key with the mcp:access scope plus all resource scopes you want the agent to use:

Navigate to Profile → API Keys in the web UI, or use the API:

POST /api/keys
{
  "api_key": {
    "label": "MCP Agent Key",
    "scopes": [
      "mcp:access",
      "playrooms:read", "playrooms:write",
      "playspecs:read", "playspecs:write",
      "playgrounds:read", "playgrounds:write",
      "playzones:read", "playzones:write",
      "repos:write",
      "launch:write",
      "import_templates:read", "import_templates:write"
    ]
  }
}

2. Configure Your AI Client

Add the following to your MCP client configuration:

{
  "mcpServers": {
    "playgrounds-dev": {
      "url": "https://fibe.gg/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_TOKEN"
      }
    }
  }
}

3. Start Using It

Ask your AI agent things like:

"List my playrooms and show which ones are active"

"Create a playspec with a Rails app, PostgreSQL, and Redis"

"Launch a playground from my web-stack playspec on playroom 1"

The agent will use the appropriate MCP tools to carry out your request.

Endpoint

Method	Path	Description
POST	/mcp	Send JSON-RPC 2.0 messages
GET	/mcp/sse	SSE stream (legacy transport)
DELETE	/mcp	Terminate session

The server supports both the Streamable HTTP transport (recommended) and the legacy SSE transport.

Authentication

• Method: Bearer token in the Authorization header
• Required scope: mcp:access on the API key
• Per-tool scopes: Each tool additionally checks resource-specific scopes (e.g., playgrounds:write)

All standard API authentication rules apply. An API key without mcp:access will be rejected with a 403 Forbidden error.

Available Tools

Playrooms (Docker Hosts)

Tool	Required Scope	Description
list_playrooms	playrooms:read	List all playrooms
get_playroom	playrooms:read	Get playroom details
create_playroom	playrooms:write	Create a new playroom
delete_playroom ⚠️	playrooms:delete	Delete a playroom
test_playroom_connection	playrooms:read	Test SSH/Docker connectivity

Playspecs (Environment Blueprints)

Tool	Required Scope	Description
list_playspecs	playspecs:read	List all playspecs
get_playspec	playspecs:read	Get playspec with services and files
create_playspec	playspecs:write	Create from Docker Compose YAML
delete_playspec ⚠️	playspecs:delete	Delete a playspec
validate_compose	playspecs:read	Validate Docker Compose YAML

Playgrounds (Running Environments)

Tool	Required Scope	Description
list_playgrounds	playgrounds:read	List all playgrounds
get_playground	playgrounds:read	Get playground details
create_playground	playgrounds:write	Create and deploy a playground
destroy_playground ⚠️	playgrounds:delete	Tear down containers
get_playground_status	playgrounds:read	Get deployment status
get_playground_logs	playgrounds:read	Get service container logs
get_playground_debug	playgrounds:read	Get diagnostic/debug info
recreate_playground ⚠️	playgrounds:write	Re-deploy from scratch
extend_playground	playgrounds:write	Extend the TTL

Playzones (GitHub Repos)

Tool	Required Scope	Description
list_playzones	playzones:read	List connected repositories
get_playzone	playzones:read	Get repository details
get_playzone_branches	playzones:read	List branches
create_playzone	playzones:write	Connect a GitHub repository

Launch

Tool	Required Scope	Description
launch_playspec	launch:write	One-shot: Docker Compose YAML → deployed environment

Templates

Tool	Required Scope	Description
list_templates	import_templates:read	List your templates
get_template	import_templates:read	Get template details and versions
search_templates	import_templates:read	Search templates by name, category, and filters
list_template_categories	import_templates:read	List available template categories
create_template	import_templates:write	Create a new template with an initial version
update_template	import_templates:write	Update template metadata (name, description, category)
delete_template ⚠️	import_templates:write	Delete a template and all its versions
list_template_versions	import_templates:read	List versions of a template
create_template_version	import_templates:write	Add a new version to a template
toggle_template_version_public	import_templates:write	Toggle a version's public/private visibility
delete_template_version ⚠️	import_templates:write	Delete a specific version
upload_template_image	import_templates:write	Upload a cover image for a template
fork_template	import_templates:write	Fork a public template into your account

Agent Data (Artefacts, Feedbacks, Mutters)

Tool	Required Scope	Description
list_artefacts	agents:read	List artefacts with optional filters
get_artefact	agents:read	Get artefact details
upload_artefact	agents:write	Upload a file as an artefact
download_artefact	agents:read	Download an artefact file
list_feedbacks	agents:read	List feedbacks with optional filters
get_feedback	agents:read	Get feedback details
create_feedback	agents:write	Create a new feedback entry
update_feedback	agents:write	Update feedback content
delete_feedback	agents:delete	Delete a feedback entry
list_agent_mutters	agents:read	List agent mutters with optional filters
get_agent_mutter	agents:read	Get mutter details
create_agent_mutter	agents:write	Create a new mutter

Advanced Filtering: The list_artefacts, list_feedbacks, and list_agent_mutters tools all support:

Parameter	Description
query	Text search across name/content
playground_id	Filter by originating workspace
sort	Sort field (e.g., created_at, updated_at)
sort_direction	asc or desc
limit	Max results to return

Webhooks

Tool	Required Scope	Description
list_webhook_endpoints	webhooks:read	List all webhook endpoints
create_webhook_endpoint	webhooks:write	Create a new endpoint
update_webhook_endpoint	webhooks:write	Update an endpoint
delete_webhook_endpoint	webhooks:delete	Delete an endpoint
list_webhook_deliveries	webhooks:read	View delivery history
test_webhook_endpoint	webhooks:write	Send a test event

Destructive Tools

Tools marked with ⚠️ have destructiveHint: true. Well-behaved AI clients will ask for confirmation before executing them.

Available Resources

Resources provide contextual information that AI agents can read:

Resource	URI	Description
Player Profile	playgrounds://player/profile	Your account info
Player Quotas	playgrounds://player/quotas	Resource quotas and current usage
API Scopes	playgrounds://api/scopes	Available and current key scopes

Security

• All requests require a valid Bearer token with mcp:access scope
• Each tool enforces resource-specific scopes (same as the REST API)
• Ownership rules are enforced — you can only manage your own resources
• Destructive tools are annotated with destructiveHint: true
• Read-only tools are annotated with readOnlyHint: true

Testing with MCP Inspector

You can test the MCP server locally using the MCP Inspector:

npx @anthropic-ai/mcp-inspector

Configure with your instance URL (e.g., https://fibe.gg/mcp) and Bearer token.