For AI agents

Let your AI agent work in PostFiber

An agent can list your channels, upload media, draft and schedule posts and read analytics in PostFiber, with the same calendar and plan limits as the dashboard. Pick the way your agent connects.

Connect over MCP

Three ways in

MCP server

A hosted server at https://www.postfiber.com/mcp. Claude and ChatGPT add it as a custom connector and sign in with OAuth; every other client sends an API key.

Agent skill

A SKILL.md that teaches an agent when and how to use PostFiber. https://www.postfiber.com/agent/SKILL.md serves it.

Guides by agent

Each guide follows the steps in the agent's own documentation. Anything that speaks MCP can follow the MCP server guide, and anything that runs shell commands can use the command line.

Agent Made by Connects with Sign-in Guide
Claude Anthropic Custom connector OAuth sign-in Open the guide
Claude Cowork Anthropic Custom connector OAuth sign-in Open the guide
ChatGPT OpenAI Custom connector OAuth sign-in Open the guide
Claude Code Anthropic MCP server API key Open the guide
Codex OpenAI MCP server API key Open the guide
Cursor Cursor MCP server API key Open the guide
Gemini CLI Google MCP server API key Open the guide
Kimi Code Moonshot AI MCP server API key Open the guide
DeepSeek Harness DeepSeek MCP server API key Open the guide
Perplexity Computer Perplexity Custom connector API key Open the guide

The skill

Save SKILL.md where your agent loads skills, or install it from https://www.postfiber.com/agent/SKILL.md if the agent can. The guides say where each agent looks.

curl -fsSL https://www.postfiber.com/agent/SKILL.md -o SKILL.md

Keep a person in the loop

An agent works with the permissions of the key, or of the person who connected it. Ask it to save drafts or schedule posts, look at the calendar before anything goes out, and give each key only the project and the permissions it needs. A key or a connected app you revoke stops working at once.

Questions

Do the API and the MCP server cost extra?
No, they are part of your plan. Plans differ in how many posts the API and the MCP server can create per day; drafts do not count.
Where do the posts an agent makes show up?
In the dashboard like any other post, in the calendar and in the post list, with its source shown.
Which networks can an agent post to?
The channels connected in the project the key or the sign-in reaches. Connect your channels in the dashboard first.
My agent is not in the list.
Any client that supports remote MCP servers can follow the MCP server guide, and any agent that runs shell commands can use the command line and the skill.
Is it safe to give an agent an API key?
Use a key limited to one project and to the permissions you need, keep it in an environment variable or in the agent's secret store and never in a prompt, and revoke it under API keys if it leaks.

MCP server guide API reference