MCP Integration

ContentForge implements the Model Context Protocol (MCP) and can be used as a tool server for AI assistants like Claude Code.

What is MCP?

MCP is a protocol that allows AI assistants to use external tools. When ContentForge runs as an MCP server, an AI assistant can create content, adapt it for platforms, schedule publication, and check analytics -- all through natural language.

Setup with Claude Code

Step 1: Start the MCP Server

Add ContentForge to your Claude Code MCP configuration. Edit your Claude Code settings (typically ~/.config/claude-code/mcp.json or the project-level .mcp.json):

{
  "mcpServers": {
    "contentforge": {
      "command": "contentforge",
      "args": ["mcp"]
    }
  }
}

Step 2: Verify

In Claude Code, the ContentForge tools will appear automatically. You can verify by asking Claude:

"What ContentForge tools are available?"

Transport Modes

ContentForge supports two MCP transport modes:

Mode	Command	Use Case
stdio	`contentforge mcp`	Claude Code, local AI tools
SSE	`contentforge mcp --sse`	Web-based AI clients

For Claude Code, stdio (the default) is the correct choice.

Available Tools

`create_content`

Create a new content piece.

Parameters:

Parameter	Type	Required	Description
`title`	string	yes	Content title
`body`	string	yes	Markdown body
`content_type`	string	yes	article, thread, short_post, etc.
`tags`	array	no	Tags for organization
`project`	string	no	Associated project name

Example prompt:

"Create a new article titled 'Understanding Rust Lifetimes' with tags rust and tutorial."

`list_content`

List content filtered by status, project, or type.

Parameters:

Parameter	Type	Required	Description
`status`	string	no	Filter by status (idea, drafting, published, etc.)
`project`	string	no	Filter by project
`content_type`	string	no	Filter by content type
`limit`	number	no	Max results (default 20)

Example prompt:

"List all my draft articles."

`get_content`

Get a specific content piece by ID.

Parameters:

Parameter	Type	Required	Description
`id`	string	yes	Content UUID

`adapt_content`

Generate a platform-specific adaptation.

Parameters:

Parameter	Type	Required	Description
`id`	string	yes	Content UUID
`platform`	string	yes	Target platform (twitter, devto, etc.)
`use_ai`	boolean	no	Use AI for intelligent adaptation

Example prompt:

"Adapt my latest article for Twitter as a thread."

`publish`

Publish content to a platform.

Parameters:

Parameter	Type	Required	Description
`id`	string	yes	Content UUID
`platform`	string	yes	Target platform

Example prompt:

"Publish the Rust lifetimes article to DEV.to."

`schedule`

Schedule content for future publication.

Parameters:

Parameter	Type	Required	Description
`id`	string	yes	Content UUID
`platform`	string	yes	Target platform
`scheduled_at`	string	yes	ISO 8601 datetime

Example prompt:

"Schedule the Twitter thread for tomorrow at 9 AM EST."

`list_platforms`

List configured platform accounts and their health status.

Parameters: None

`get_analytics`

Get engagement metrics for published content.

Parameters:

Parameter	Type	Required	Description
`id`	string	no	Content UUID (omit for all)

Example Conversations

Full Publishing Workflow

You: "Write a blog post about why developers should use Rust for CLI tools, then publish it to DEV.to and create a Twitter thread for it."

Claude: Uses create_content to draft the article, then adapt_content for DEV.to and Twitter, then publish to both platforms. Reports the live URLs.

Scheduled Campaign

You: "I have three articles ready. Schedule them for this week -- one on Monday, Wednesday, and Friday at 9 AM on LinkedIn and Twitter."

Claude: Uses list_content to find ready articles, then schedule for each platform and date.

Analytics Check

You: "How did my posts perform this week?"

Claude: Uses get_analytics and summarizes views, likes, and engagement across platforms.

Security Considerations

MCP over stdio communicates only with the parent process (Claude Code). No network exposure.
Platform credentials are read from your local config. The MCP server does not accept credentials as tool parameters.
All publish actions are explicit -- the AI must call the publish tool; no automatic publishing happens.
Audit: All operations are logged via the standard tracing framework. Set RUST_LOG=contentforge=info to see MCP tool invocations.