Launching with MCP server

Postnavo MCP server

The Model Context Protocol (MCP) is the open standard AI assistants use to connect to apps. Postnavo is launching with a hosted MCP server with OAuth sign-in, so any MCP-compatible agent can schedule posts on your behalf.

MCP client config

{
  "mcpServers": {
    "postnavo": {
      "url": "https://mcp.postnavo.com"
    }
  }
}

Illustrative — final endpoints and commands are published at launch.

What your agent can do

List connected channels
Create, schedule and update posts
Upload images
Read the calendar and queue
Fetch post analytics
Require human approval before publishing

Works with

How a hosted MCP server works

What MCP is, in the protocol’s own words

The protocol’s site describes MCP as “an open-source standard for connecting AI applications to external systems”, and compares it to a USB-C port for AI applications.[1]

An MCP server offers tools. An AI application, the client, connects to the server and calls them. Postnavo is launching with a server whose tools are the scheduler itself: list your channels, create and schedule posts, read the calendar.

A hosted server is an address, not a program on your computer

The specification has two standard transports. With stdio, messages travel over the standard streams of a subprocess that the client launches. With Streamable HTTP, each message is an HTTP POST to a single MCP endpoint.[2]

Postnavo is launching with the second kind: a hosted server that a client reaches at an address. You install nothing, and it suits the chat apps, which take a server URL.

A hosted server is called from the vendor’s side, so it has to be reachable from the public internet. Claude’s and Grok’s docs both say so about the servers they connect to.[8][14]

Sign-in is OAuth, so there is no key to paste

In the specification authorization is optional. Where an HTTP server supports it, the flow is OAuth 2.1: the MCP server acts as a resource server, and the client makes requests on behalf of the person who signed in.[3]

Postnavo is launching with OAuth sign-in on its hosted server. You approve access in your browser, and no API key is copied into a config file.

Anthropic describes the effect for its own client: Claude can act for you “without Claude ever seeing your actual password”, and you can revoke the permission by disconnecting the connector.[8]

The protocol asks for a person in the loop

On tools the specification is direct: “there SHOULD always be a human in the loop with the ability to deny tool invocations”.[4]

A server can describe a tool with hints such as readOnlyHint and destructiveHint. Clients must treat those annotations as untrusted unless they come from a trusted server.[4][5]

Every client turns that into its own prompt, and several let you switch the prompt off for a tool you trust. That is why Postnavo is launching with an approval setting of its own: while it is on, a post an agent creates waits for your OK, whatever the client asked.

Where each agent takes a server address

ChatGPT
ChatGPT Plugins, the plus button, then “Add custom MCP server”. Writing through a custom server is in beta on Business, Enterprise and Edu.[6][7]
Claude
Customize, Connectors, “+ Add”, then “Add custom connector”. On every plan; Free is limited to one custom connector.[8]
Claude Code
claude mcp add --transport http <name> <url>, then /mcp to sign in.[9]
OpenClaw
openclaw mcp add <name> --url <url>, or Settings, MCP in the Control UI.[10]
Codex
codex mcp add <name> --url <url>, then codex mcp login <name>.[11]
Cursor
A named entry with a url under mcpServers in .cursor/mcp.json.[12]
Gemini
Settings, Connected Apps, “Custom apps”, in the web app. For people in the US on a personal Google Account.[13]
Grok
grok.com/connectors, “New Connector”, then “Custom”.[14]

FAQ

Frequently asked questions

What is MCP?

The Model Context Protocol is an open standard that lets AI applications connect to tools in other apps. Postnavo is launching with an MCP server, so an agent can schedule posts for you.

Do I need an API key?

No. The hosted MCP server is launching with OAuth sign-in, so you approve access in your browser.

Does the Postnavo MCP server run on my computer?

No. It is launching as a hosted server that your agent reaches at an address, so there is nothing to install or keep running.

Which AI apps can add a remote MCP server?

ChatGPT, Claude, Claude Code, OpenClaw, Codex, Cursor, Gemini and Grok each document a way to add one, with their own plan and admin rules. The list on this page says where, as read on 2026-10-12. Postnavo’s server is launching soon.

Can I limit what an agent does through MCP?

The protocol expects a person to be able to deny a tool call, and clients ask before a tool runs unless you tell them not to. Postnavo is launching with an approval setting that holds agent-made posts until you OK them.

Sources, checked 2026-10-12

Blog

From the blog

Read all posts

Get your weekends back. Let AI handle your social.

Join the waitlist and lock in 30% off for life.

Founding price for life
No credit card
Unsubscribe anytime