The MCP server: how an AI agent schedules a post on Timalio
Timalio ships a Model Context Protocol (MCP) server with eight tools: including one that serves each network’s current posting rules: so an AI agent can schedule a post the same way a person does, and it still has to be approved before it publishes. Among the tools in this market, only Postly offers something comparable.
Why "current rules", not a training cutoff, matters
A language model’s knowledge of a network’s character limit or image count is frozen at its training date. Timalio’s `network_rules` tool answers instead from a table that is updated when the rule changes, and marks each fact officiel (documented by the network) or observed (seen in practice, not officially documented): so an agent knows how much to trust each constraint.
Connecting an agent
{
"mcpServers": {
"timalio": {
"url": "https://app.timalio.com/api/mcp?names=en",
"headers": { "Authorization": "Bearer tml_YOUR_KEY" }
}
}
}HTTP transport, one API key. The same key also works on the plain REST API: see [the full reference](sujet:api).
The eight tools: read a network’s current rules, list accounts, list and read posts, schedule a post, act on a post (submit, approve, request changes…), upload an image, retry a first comment.
Still reviewed, always
A post scheduled by an agent is born a draft, exactly like one typed by hand. It still has to be submitted and approved before it publishes: see how approval works. An AI that publishes without review is exactly what Timalio does not sell.
Frequently asked questions
Do other scheduling tools offer an MCP server?
Among the tools compared on the Buffer alternative page, only Postly advertises an MCP server; the rest document a REST API at best.
Can an AI agent publish a post without a human reviewing it?
No. Every post created through the MCP server is born a draft and follows the same submit-and-approve loop as a post typed by a person.