Limits

Hard caps

Limit	Value
Ideas per POST /ideas/import call	25
Ref window returned by GET /ideas/mining-context	90 days
Brand header	X-Postlyra-Brand

Tool	Cap
push_ideas	25 ideas per call
list_ideas	50 results, no pagination cursor
get_mining_context	Refs from the last 90 days

list_ideas returning has_more: true means there are more than 50 matching ideas and you cannot page past them. Narrow with the status or format arguments instead.

Rate limits

There is no rate limit on the MCP server today. The OAuth endpoints behind connector setup are limited, but a client only touches those while connecting and while refreshing a token, so you will not meet them.

That is not a guarantee

It is a current implementation fact. Limits get added when they are needed, and a client that hammers the server will be the one that breaks first. Batch pushes into calls of 25, and do not poll list_ideas in a loop.

Plan allowances

MCP calls do not consume AI credits. Your agent's own model calls are billed by whoever provides that model, not by Postlyra, which is the whole point of mining client side.

What plan limits do apply: the brands you can reach are the brands your account has.

What is not supported

	Status
Per-brand connections	Not a thing. One connection covers your whole account, by design
Read-only API tokens	Not implemented. An API token carries full access. Scopes exist for OAuth connections only
Approving or scheduling without your consent	Impossible: approve_post and create_post's approve argument both need the elevated access level, which you consent to explicitly. Standard access can only add to your review queue
Publishing outside the scheduled tick	Not possible from MCP at all, even with elevated access. Approving only makes a scheduled publication eligible; the same publish:tick that fires Studio-scheduled posts is what sends it
Reading your local files	The server never does this. Your agent does