The social media MCP server that proposes — not posts.
Connect Claude, or any MCP agent, to your social accounts through Skedlark. It reads your real analytics, drafts posts in your voice, and queues a schedule across Threads, LinkedIn, Pinterest, Reddit, and Google Business. Then it stops — because nothing publishes until you approve it.
read + propose scopes only·approval queue·EU-hosted
What is a social media MCP server?
The Model Context Protocol (MCP) is an open standard for giving an AI agent a set of tools it can call. A social media MCP server exposes your social accounts to the agent as those tools — so the agent can read your analytics, draft a post, or queue a schedule without you copy-pasting between a chat window and a dashboard.
Most tools in this space hand the agent a publish tool: the agent is trusted to post for you. Skedlark makes a different choice on purpose. The tools it exposes can read and propose — never publish. You stay the one who presses go.
7 tools, granted read and propose scopes.
The agent gets exactly enough to be useful and nothing more. Here is the shape of what it can — and cannot — do.
Read your numbers
Pull cross-channel analytics, top posts, and audience signal — rendered as an LLM-ready brief the agent can actually reason over.
Draft in your voice
Generate posts grounded in the brand context you locked, under a strict anti-fabrication rule: it may only cite proof you supplied.
Queue a schedule
Propose times and a per-channel plan across Threads, LinkedIn, Pinterest, Reddit, and Google Business — as a proposal, not a publish.
Check the queue
Read the approval queue and the audit trail so the agent knows what is pending, what you approved, and what already shipped.
There is no publish tool in the default scope set. Auto-publish is off unless you deliberately enable it for a specific channel.
Capture, draft, propose — you approve.
Capture the idea
You hand the agent a thought — or it works from a brief built out of your own analytics.
Draft, grounded
It writes posts in your locked voice, citing only numbers you actually gave it.
Propose a schedule
It queues a per-channel plan and drops every draft into your approval queue.
You approve
You review, edit, and approve. Only then is anything scheduled to post.
An agent that can post is an agent that can post the wrong thing, at the wrong time, as you.
Give an agent a publish scope and you are one hallucination away from a live mistake in your name. Skedlark keeps the publish decision human: read and propose scopes, an approval queue, and a full audit trail of every action. Even a compromised agent cannot silently post as you.
Where the agent can propose to post.
Skedlark is deliberately focused on the channels where consistent, grounded posting compounds — not a long list you will never connect.
Each has a grounded growth playbook in the Skedlark Academy.
Frequently asked
- What is a social media MCP server?
- An MCP (Model Context Protocol) server is a standard way to let an AI agent — like Claude — use an external tool. A social media MCP server exposes your social accounts to the agent as a set of tools it can call: read analytics, draft a post, queue a schedule. Skedlark is one such server, with a deliberate constraint: the tools it exposes can read and propose, but they cannot publish on their own.
- Can Claude post to my social media through Skedlark?
- Claude can draft posts and propose a schedule for you, but it cannot publish. Every draft an agent creates lands in an approval queue. Nothing goes live to Threads, LinkedIn, Pinterest, Reddit, or Google Business until you approve it. The agent is granted read and propose scopes only — never a publish scope by default.
- Which AI agents work with Skedlark?
- Any MCP-compatible client. Skedlark is built for Claude first, but the Model Context Protocol is an open standard, so an agent running in any MCP-aware environment can connect. You install the Skedlark MCP server, authorize the scopes, and the agent gains its 7 tools.
- Does the agent post automatically on a schedule?
- No. Auto-publish is off by default. The agent proposes a schedule; you approve it; only then does it enter the queue to be posted at the chosen time. Direct-publish — where an approved channel posts without a manual click — is an option you deliberately turn on for a specific channel on higher tiers, not a default.
- How is this different from a scheduler with a posting API?
- A posting API lets an agent publish for you — the agent is trusted to act. Skedlark inverts that: the agent is trusted only to read and to propose, and you are the one who acts. That governance boundary — read + propose scopes, an approval queue, and a full audit trail of every action — is the whole point. It means even a misbehaving or compromised agent cannot silently post as you.
- Is my data safe with an agent connected?
- Skedlark is EU-hosted and GDPR-native. Connections to your social accounts use each platform's OAuth, so we never see your passwords, and the tokens are encrypted at rest with AES-256-GCM. Every action the agent takes — read, draft, schedule — is written to an audit trail you can review. Drafts are also grounded in an anti-fabrication rule: the agent may only cite analytics you actually supplied.
Give your agent the keys to draft — and keep the keys to publish.
Skedlark is in founding beta. The first 50 founders get the full Standard plan for $5/mo, locked for six months.