What is MCP?
The Model Context Protocol (MCP) is an open standard that lets AI agents — assistants running in your editor, terminal, or chat client — call structured "tools" against a service, instead of guessing at a web UI or scraping HTML. An MCP server exposes a fixed set of tools (with typed parameters and descriptions); an MCP client (your agent) discovers those tools and calls them on your behalf, inside a conversation.
Postnomic ships a hosted MCP server so an agent can manage your blog content and read your analytics directly, using the same actions you'd otherwise perform by hand in the dashboard.
The Postnomic MCP server
The server is hosted at:
https://mcp.postnomic.com/mcp
It speaks MCP over Streamable HTTP, the standard transport for remote MCP servers — your client connects to that single URL rather than launching a local process.
A thin adapter, not a separate system
The MCP server is a thin adapter over the same REST API that powers the dashboard, the Client SDK, and every other Postnomic integration. It does not talk to the database directly and it does not have its own notion of permissions. Instead:
- Every tool call is authenticated with your Personal Access Token (PAT).
- The server forwards your token's
Authorizationheader unchanged to the API. - The API runs the request exactly as if you had made it — with your existing blog roles (Admin, Author, etc.) and your account's plan limits.
In practice this means the agent can never do more than you can. If you're not an Author on a blog, the agent can't publish to it either. If your plan has hit its monthly post quota, tool calls fail with the same quota error the dashboard would show.
MCP access requires a paid plan. Free-tier accounts can use the dashboard and API as usual, but the MCP server itself is a Plus/Pro/Enterprise feature. Free-tier tokens are rejected by the server with an access-denied response.
What you can do
Once connected, an agent can:
- Author and publish posts — create drafts, update titles/content/tags/categories, publish, unpublish, schedule for a future date, or restore an archived post to draft.
- Manage translations — add or remove a post's content in one of your blog's other languages.
- Upload images — send image bytes to a blog's media library and get back a URL to use as a post's cover, thumbnail, or share image.
- Read analytics — pull a blog's traffic dashboard or a single post's statistics, so an agent can summarize performance without you opening the Analytics page yourself.
See the MCP Tools Reference article for the full list of tools, their parameters, and how they're grouped.
Getting connected
How you connect depends on which client you're using:
- Local agents (editors, terminals, Claude Code, custom MCP clients) — connect with a Personal Access Token, sent as a bearer token. See Connect a Local Agent (Personal Access Token).
- claude.ai (web, desktop, mobile) — add Postnomic as a custom connector using OAuth; no token to copy or store. See Add the claude.ai Connector.
- Microsoft Copilot / Microsoft 365 — connect Postnomic as a custom connector inside Copilot. See GitHub Copilot, Microsoft 365 & Other Agents.
All three paths reach the same server, run through the same API, and enforce the same permissions and plan limits — pick whichever matches the client you're using.
Open source
The MCP server itself is part of the Postnomic platform and isn't open source. The .NET Client SDK — a separate, optional set of NuGet packages developers use to embed a Postnomic blog into their own ASP.NET Core or Blazor application — is MIT-licensed and open source at:
https://github.com/threeb-it/postnomic-dotnet
The Client SDK and the MCP server are unrelated integration paths: the SDK renders blog content inside a host application, while the MCP server lets an AI agent manage that content through conversational tool calls.