The social scheduler built for humans and their agents.
Capture ideas from anywhere, let your agent draft from your real performance data, and approve from one queue. Five channels live today. Full control, always.
Works with Claude Desktop · Claude Code · Cursor · any MCP client
// agent calls the Skedlark MCP server
▸ schedule_post({
channel: "threads",
text: "Shipped the approval queue today…",
at: "tue 09:30"
})
// Skedlark responds — it does NOT publish
◂ { status: "pending_approval",
queue: "/approvals/8f3a",
audit: "logged" }You want your AI to actually run your social media. Today you pick your poison: autopilot tools that post without asking, or dashboards where you still do everything by hand.
One you can't trust. The other isn't hands-off. Skedlark is the third option.
Ninety seconds. Capture, draft, approve, publish.
The whole loop, in the real app — no slides, no mockups.
Four steps. The agent does three of them.
This loop is the whole product — and the one thing a send-button MCP can't draw.
Capture
Drop an idea from your phone, or talk to your agent. It lands in your inbox.
Brief
Your analytics become an LLM-ready context block — what worked, when, for whom.
Propose
The agent drafts within scoped permissions. It can suggest; it can't publish.
Approve
Review in one queue, approve, and Skedlark publishes — with an audit trail.
It's a real app, not an API you script blind.
Most agent-native tools are just an API — you fire commands into the dark and hope. Skedlark is the bridge: the agent does the work, and you get a real dashboard to see it.
- A visual calendar of everything queued across channels
- Cross-channel analytics you can actually open and read
- An approval queue where you check every draft before it ships
No living in the terminal. The same data your agent drafts from is right there for you to analyze and decide on yourself.
Agent power, with the brakes others skip.
Works with the agents you already use
An MCP server with 13 tools. Connect Claude Desktop, Claude Code, or Cursor — or capture a post idea from your phone and let your agent pick it up. One-line config, copy-paste install.
Nothing ships without you
Agents get read + propose scopes, never silent publish. Every draft lands in an approval queue with a full audit trail. Your brand voice is locked by you — by design, no agent can rewrite it.
Grounded in your real numbers
Cross-channel analytics render into LLM-ready briefs, so drafts cite what actually worked. Anti-fabrication is the rule: the AI may only use proof points you supplied. Every other AI tool hallucinates your wins — Skedlark only uses proof you actually have.
What can the agent actually do?
The honest answer is the feature. Agents read and propose. They never publish on their own, and they can never rewrite the brand voice you locked. Every action is logged.
Hosted in the EU (Hetzner). OAuth tokens encrypted at rest with AES-256-GCM. Read the security page →
| What an agent can do | Allowed? | Scope |
|---|---|---|
| Read your analytics & posts | Yes | read scope |
| Draft & propose a post | Yes | propose scope |
| Schedule into the approval queue | Yes | propose scope |
| Publish without your approval | Never | blocked by design |
| Rewrite your locked brand voice | Never | blocked by design |
Five channels live today. We only promise what works.
Threads
Text-first posts, queued with everything else.
Personal & company pages.
Pins on a schedule.
Per-subreddit timing.
Google Business
Local posts that show up.
X / Twitter
On the roadmap.
One product, three ways to live in it.
Post while you build. Capture an idea mid-commit from your phone, let your agent turn it into a thread grounded in what's already landing, and approve it from one screen before bed.
The two rows nobody else can check.
Plenty of tools connect to an agent now. None of them let the agent work and keep you in control with a real app behind it.
| Feature | Skedlark | Postiz | Buffer | Raw API |
|---|---|---|---|---|
| Connect any AI agent (MCP) | ||||
| Agents propose, you approve | ||||
| Real dashboard: calendar + analytics | partial | |||
| Anti-fabrication brand context | ||||
| Cross-channel analytics as agent briefs | partial | |||
| EU-hosted by default | partial | |||
| No per-channel price creep | — |
Postiz ships no approval workflow (a cited user complaint); Buffer is US-hosted and priced per channel; raw social APIs (Ayrshare-style) have no dashboard at all. Public info as of June 2026.
Paid from day one. No free tier, no leeches.
The first 50 founding members claim a seat for $15 — three months at $5/mo, paid up front — and lock that founding rate.
Founding 50 · $15
Three months at $5/mo up front, then your founding rate stays locked.
Workspace-based
Not per-channel. Add all five channels without the bill climbing.
Agent usage included
No credit metering. Your agent drafts as much as it needs.
Why I built Skedlark, and how I use it — from the founder.
The questions skeptics ask first.
Let your agent run social. Keep your hand on the switch.
Claim one of 50 founding seats — $15 for three months at $5/mo, locked for good.