Mutiny MCP Overview
Mutiny connects to AI assistants through the Model Context Protocol (MCP), an open standard that lets AI apps like Claude and ChatGPT call external tools. With MCP, your AI assistant can create, manage, and publish customer-facing assets in your Mutiny workspace on your behalf.
What is MCP?
MCP is an open protocol that standardizes how AI assistants connect to external tools. Instead of custom integrations per platform, MCP provides a universal interface. Any compatible AI client can discover and call tools from any MCP server.
How Mutiny Uses MCP
Mutiny runs a remote MCP server that exposes your workspace’s asset creation, content library, and publishing capabilities as tools. When you authorize a connection, the AI assistant receives a scoped OAuth token that determines what it can do in your workspace.
This means you can research a target account, draft a personalized proposal, upload brand assets, and publish, all from a single conversation, without switching to the Mutiny app.
Supported Clients
Mutiny will be listed in the Claude Connectors directory and ChatGPT Apps marketplace soon. Until then, use the custom connector setup instructions below.
Claude
- Claude web (claude.ai ): via the Connectors directory
- Claude Desktop (macOS & Windows): via Connectors
- Claude Code (CLI): via MCP server configuration
→ See Connecting Mutiny to Claude
ChatGPT
- ChatGPT web (chatgpt.com ): via the Apps marketplace
- ChatGPT Desktop (macOS & Windows): via Apps
→ See Connecting Mutiny to ChatGPT
Other Clients
Mutiny’s MCP server works with any client that supports remote MCP servers over HTTP with OAuth authentication, including IDE integrations, custom agents, and other AI tools.
You can use the Mutiny MCP server URL to set up the Mutiny MCP:
What Can You Do?
- Create landing pages, proposals, pitch decks, case studies, meeting recaps, and more
- Combine account research and CRM data already in your conversation with Mutiny’s asset creation
- Browse templates, manage your content library, and publish, all without leaving your AI assistant
Tools
The MCP server exposes 12 tools across three categories. All tools include proper readOnlyHint, destructiveHint, idempotentHint, and openWorldHint annotations. One tool (get_task_status) is app-only and hidden from external clients. It backs the interactive asset preview widget’s polling flow.
Asset Management (7 tools)
| Tool Name | Description | Read/Write |
|---|---|---|
list_assets | List existing assets with optional filters for status and creator (defaults to assets created by the current user) | Read |
get_asset | Retrieve a specific asset with links, screenshot, and all-time analytics (views/clicks) | Read |
get_assets_analytics | Get ranged analytics (views/clicks) for multiple assets over a start/end timestamp window | Read |
create_asset | Create an asset from a template, blueprint, or from scratch and queue the initial agent prompt as a workflow run | Write |
send_agent_message | Send a refinement message to an existing asset agent and queue a new workflow run for iterative revisions | Write |
publish_asset | Publish an asset to generate its live/shareable link | Write |
get_task_status (app-only) | Returns workflow run status for the interactive asset preview widget’s polling flow. Hidden from external clients. | Read |
Templates (1 tool)
| Tool Name | Description | Read/Write |
|---|---|---|
list_templates | List the company’s approved templates for choosing a starting point before asset creation (supports search query) | Read |
Content Library Management (4 tools)
| Tool Name | Description | Read/Write |
|---|---|---|
list_library_content | Search library content with filters (query, tags, type, sort) to gather context sources for asset creation or refinement | Read |
create_library_content | Upload content to library from URLs | Write |
update_library_content | Update metadata (title, description, tags) | Write |
delete_library_content | Remove library content | Write (destructive) |