An MCP server for SEO and content distribution.
Connect Parasite SEO Poster to Claude, Claude Code, or any other MCP client, and drive the whole pipeline from a conversation: point it at a page, and it breaks the page into distinct content angles, writes a native post for every social channel you have connected, and publishes them on your say-so. Every post links back to the page it came from.
https://www.parasiteseoposter.com/api/v1/mcp
Install the plugin
The plugin bundles the MCP connection with the workflow knowledge around it - what order the calls go in, which ones cost money, and why an image has to be attached before you generate. Claude Code prompts for your API key on install.
$ /plugin marketplace add bryandoreian/parasiteseoposter.com$ /plugin install parasite-seo-poster@parasite-seo-poster
You get three commands - /parasite-seo-poster:distribute to take a URL through to reviewable drafts, :drafts to review what is queued, and :check-setup to verify the connection without spending anything.
Or connect the server directly
It is a remote Streamable HTTP server, so any compliant client can reach it with an API key from Settings → API. Use the www. host: the bare domain answers with a redirect that many clients drop the request body through.
$ claude mcp add --transport http parasiteseoposter https://www.parasiteseoposter.com/api/v1/mcp \$ --header "Authorization: Bearer psp_live_..."
Discovery document: https://www.parasiteseoposter.com/mcp.json. Full endpoint reference, error codes and rate limits are in the API and MCP guide.
Six tools, and what each one costs
Reads are free and rate-limited to 60 requests a minute. The metered calls check your balance beforedoing any work, so a call you can't afford is refused rather than half-run.
List the workspace's generated posts. Free (rate-limited, not billed). Optional filters: domain_id, platform, status, limit.
Generate draft posts for an atomized page — one per connected channel of the page's domain. Metered against the workspace's API credit balance; costs nothing when every channel is already covered.
Fetch a URL, extract content atoms with the LLM, and store the page. The URL's domain must already exist in the workspace. Metered against the API credit balance.
Search Pexels and Unsplash for a photo (the same library the dashboard's picker uses). Free. Pass a result straight to attach_image.
Attach an image to a page so generated posts carry it. Free. Use image_url for your own image, or source_url from a search_images result. Attach BEFORE generate — Pinterest cannot publish a post with no image.
Publish a generated post now, or schedule it with scheduled_at (ISO-8601, future). Metered; an X/Twitter publish additionally debits the X credit wallet, same as from the dashboard.
An agent gets the same review queue you do
Generation stops at drafts. Nothing reaches a social account until a publish call runs, and that tool is the one marked destructive in the catalog, which is what makes a client stop and ask before it fires. The key acts as the whole workspace, so it is minted by an admin, shown once, and revocable from Settings → API at any time.
The plan limits apply unchanged: a free workspace has no connected channels, so there is structurally nothing for an agent to publish to until you connect one.
Common questions
What is an MCP server, and why does this one exist?
MCP (Model Context Protocol) is the standard way an AI assistant connects to an outside tool. This server exposes the same pipeline the dashboard runs - atomize, generate, publish - so you can drive distribution from a conversation, or from an agent, instead of clicking through the app.
Does the MCP server do anything the API doesn't?
No, deliberately. Every tool call is forwarded to the same /api/v1 endpoint a curl would hit, carrying the same key, through the same auth, rate limiting and metering. MCP is a thin adapter over the REST API, not a second path with its own rules.
Will it publish something without asking me?
It can't publish anything you haven't set up: publishing requires a paid plan with a connected channel, and the tool that publishes is flagged destructive, so clients confirm before running it. Generation stops at drafts, which land in the same review queue the dashboard uses.
What does it cost to run?
Reads are free. Atomizing a page is $0.40, generating is $0.15 per post actually created, and publishing is $0.05 per post. The balance is checked before any work runs, so an out-of-credits call is refused and costs nothing.
Which clients does it work with?
Any MCP client that speaks Streamable HTTP and can send an Authorization header - Claude Code, Claude Desktop, and the growing set of third-party agents and IDEs. The Claude Code plugin is a packaged version of the same connection.
Is automated distribution against search engine guidelines?
Publishing original, platform-native content with a contextual link to your own page is normal distribution. What draws penalties is fabricated engagement, fake personas, and the same text duplicated everywhere - none of which this does. Each post is generated from a distinct content angle.