MCP Server
Connect FeedSquad to Claude, ChatGPT, Cursor, or any MCP-compatible AI client. Post to LinkedIn, X, and Threads from chat — with OAuth, scheduling, cadence enforcement, and per-platform adaptation.
Install in 60 seconds
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"feedsquad": {
"url": "https://feedsquad.com/api/mcp"
}
}
}Claude Code / Cowork
claude mcp add feedsquad https://feedsquad.com/api/mcp
ChatGPT
Settings → Connected Apps → search FeedSquad → Connect.
Cursor & other MCP clients
Any client that supports the MCP HTTP transport. Point it at https://feedsquad.com/api/mcp.
What it can do
Twenty tools, grouped into four areas. Every tool returns structured responses with recovery_hint fields on errors so agents can self-correct.
Drafting & scheduling
create_post· idempotent drafts with optional schedulingupdate_post·reschedule_post·delete_postset_post_image· attach validated HTTPS imagesconfirm_post·publish_post· idempotent publish path
Campaigns
create_campaign· batch cadence pre-flight, no silent breachesconfirm_campaign· approve all in one calladapt_for_platform· LinkedIn ↔ X ↔ Threads adaptation
Discovery & analytics
get_calendar· filter by platform, status, dateget_calendar_health· gaps, clusters, cadence driftget_post·search_postsget_performance_insights· metrics + AI summary
Account & billing
connect_platform· OAuth to LinkedIn / X / Threadslist_accounts· personal + LinkedIn company pagesget_usage·get_plans·get_help
Built-in safeguards
Three problems agents typically run into are solved at the server layer — so the LLM doesn’t have to.
Cadence pre-flight
Per-platform daily / weekly caps + minimum spacing are enforced when scheduling, not at publish time. Timezone-aware so a Tokyo founder doesn’t silently bypass the cap by straddling UTC midnight.
Idempotent retries
Pass an idempotency_key on create_post or publish_post. Retries after a network drop replay the original row instead of creating duplicates.
Recovery hints
Every error response includes a recovery_hint pointing at the next tool to call (with pre-filled args where possible) — so agents stop retrying blindly.
Cookbook & reference
Common recipes
Scopes
OAuth 2.1 with PKCE. Tokens are bound to https://feedsquad.com/api/mcp per RFC 8707.
posts:readposts:writeposts:publishcampaigns:writeconnections:writeanalytics:readbilling:readSecurity
OAuth 2.1 with PKCE — credentials never reach the AI provider.
Approval-gated publish — drafts never go live without an explicit confirm step.
Scoped permissions — agents only get the scopes they need; revoke from Settings.
Request IDs — every response includes x-mcp-request-id for support correlation.
FAQ
What is MCP?
The Model Context Protocol is an open standard that lets AI assistants connect to external tools. FeedSquad implements MCP so you can manage social media from your AI chat.
Do I need API keys?
No. FeedSquad handles all platform integration. You connect social accounts via OAuth — no developer app, no API keys.
Is there a free tier?
Yes. Handler Free gives you 10 posts per month with the calendar and campaign tracking. See pricing for upgrade options.
Can I see what changed?
Yes. Every action runs through the same create_post → confirm_post → publish_post flow visible in your calendar. Agents can use dry_run: true on most write tools to preview the effect without committing.
Ready to connect?
Free to start — install in your AI client of choice, sign in with FeedSquad, and post.