Engineering note
Let Claude post for you: Eurekuh’s social media MCP server
Eurekuh’s MCP server lets AI agents like Claude, ChatGPT and Cursor schedule and publish posts to ten social networks. Seven tools, Bearer-key auth, confirmation before posting, free on every plan.
AI assistants can write a week of posts in a minute. Getting those posts onto TikTok, LinkedIn and Bluesky still takes a person copying and pasting. Eurekuh closes that gap with a Model Context Protocol (MCP) server: connect it to Claude or another agent and the agent can check your accounts, validate captions and schedule posts across ten networks.
What the MCP server can do
The server exposes seven tools:
- list_accounts: see which social accounts are connected.
- check_limits: validate a caption against each network’s limits before committing to it.
- upload_media: attach images or video.
- create_post: draft, schedule or publish a post.
- list_posts: review what is queued and what has gone out.
- get_post_status: check whether a post published on each network.
- cancel_post: pull a scheduled post.
create_post and cancel_post ask for confirmation before they act, so an agent cannot publish or delete anything without you approving it.
How to connect it
The server runs at https://eurekuh.com/api/mcp over Streamable HTTP and authenticates with a Bearer API key from your Eurekuh account. It is free on every plan, including the free one. Eurekuh has setup guides for Claude, Claude Code, ChatGPT, Cursor and Gemini.
A typical agent workflow
- Ask the agent to turn a blog post or product update into posts for each network.
- It calls list_accounts to see where you can post, then check_limits on each caption, rewriting anything over a network’s limit, such as X’s 280 characters or Bluesky’s 300.
- It calls create_post with a schedule, and you confirm.
- Later, get_post_status tells you what went out.
Why we built it this way
At Olympus ML we build agentic systems for clients, and the same rules apply here: give the model narrow tools with clear names, validate before acting, and keep a human confirmation on anything irreversible. A check_limits tool lets the agent fix its own caption before a post fails on publish, which is cheaper than handling an error from a network API afterwards.
For developers who prefer plain HTTP, paid plans also include a REST API at /api/v1, documented in an OpenAPI spec. More on agents: Eurekuh for AI agents and the Eurekuh MCP page.
Want an MCP server or agent workflow for your own product? Talk to us.