A social media MCP server
One connection that lets an agent draft, check and schedule a post across ten networks — instead of ten sets of API credentials, ten sets of rules, and ten different ways to fail at publish time.
Not available yet. The MCP server is in development. It is not available yet, and nothing on this page will connect to it today.
What works right now
An agent can already read the whole product in one fetch. This is a real file, not a summary — every caption limit, image count and platform constraint in it is generated from the same code that enforces them at publish time.
https://eurekuh.com/llms-full.txtPaste that into Claude or ChatGPT and ask it something specific — how many images Bluesky takes, whether a personal Facebook profile can be scheduled to, what the Threads limit is. It will answer from the file. There is also a shorter index at /llms.txt.
What the server will expose
Six tools, scoped to what the publishing pipeline already does today. Nothing here is a new product — it is an entry point onto the queue that runs every scheduled post now.
- list_accounts
- Which networks are connected, and which account on each.
- create_post
- A draft: caption, media, the networks to publish to, per-network overrides.
- schedule_post
- Put a draft in the queue for a given time, or publish it now.
- check_limits
- Whether a caption and media pass every selected network's rules — before anything uploads. The same check the composer runs.
- get_post_status
- Whether a scheduled post went out, and the platform's own error if it did not.
- list_scheduled
- What is in the queue.
check_limits is the one worth calling out. An agent writing a caption has no idea that Bluesky counts graphemes rather than characters, or that Pinterest takes one image and Instagram takes ten. Handing it the rules before it commits is the difference between a post that goes out and a queue full of failures at 9am.
The networks
TikTok, Instagram, YouTube, LinkedIn, Threads, Facebook, Bluesky, Google Business, Pinterest, X. All published natively through each platform's official API — the same path the composer uses, with the same limits enforced.
Per-client notes
Questions
Can I connect an agent to Eurekuh today?+
No. The MCP server is in development and there is no endpoint to point a client at. What does work today is reading: eurekuh.com/llms-full.txt describes every network, limit and constraint in one file, and any agent with a fetch tool can use it.
Why build an MCP server rather than just an API?+
Both, in that order — the server is a surface over the API, not a replacement for it. An API is what you script against; MCP is what lets someone connect Eurekuh to Claude or Cursor without writing any code at all.
Will an agent be able to post without me seeing it first?+
That is the design question that matters most and it is not settled. Publishing on your behalf, unattended, to ten public accounts is a large amount of trust to hand to a model. Expect scoped tokens and a draft-then-approve default rather than publish-immediately.
Does anyone else offer this?+
Yes — Post Bridge and Blotato both ship agent access today, and we would rather say so than pretend the category is empty. If you need it working this week, they have it and we do not.
Ready to start publishing?
Connect your accounts, upload a video, and publish. Takes less than a minute. Free to start — see plans.