For AI agents
Give your AI agent a social media workspace
Aback Social ships a Model Context Protocol server. An agent that can talk to MCP can read your calendar, draft posts, schedule them, request approval and publish them, using the same permissions and the same approval rules as the product.
How it works
Three steps take a workspace from a signed in user to a working agent.
Create an API key
Open Settings, API keys and create a key. Choose the scopes the agent needs. A key with no explicit scope is read only.
Copy the key once
The plaintext key is shown exactly once after creation. Only a hash is stored, so a lost key cannot be recovered and has to be replaced.
Point your client at the endpoint
Add the endpoint and the bearer header to your MCP client. The server speaks the Streamable HTTP transport with JSON-RPC 2.0 messages.
Connect a client
The endpoint is a Streamable HTTP MCP server. Every exchange is a JSON-RPC 2.0 message, so any client that supports remote MCP servers works.
Claude Desktop
Add the server to claude_desktop_config.json.
{
"mcpServers": {
"aback-social": {
"url": "https://abacksocial.com/api/mcp",
"headers": {
"Authorization": "Bearer absk_YOUR_API_KEY"
}
}
}
}Claude Code
Register the remote server from the terminal.
claude mcp add --transport http aback-social https://abacksocial.com/api/mcp \
--header "Authorization: Bearer absk_YOUR_API_KEY"Cursor
Add the server to .cursor/mcp.json in your project.
{
"mcpServers": {
"aback-social": {
"url": "https://abacksocial.com/api/mcp",
"headers": {
"Authorization": "Bearer absk_YOUR_API_KEY"
}
}
}
}Any HTTP client
Call the endpoint with a JSON-RPC 2.0 message.
curl -X POST https://abacksocial.com/api/mcp \
-H "Authorization: Bearer absk_YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'Available tools
The server exposes 12 tools. Each one returns structured content and reports domain failures as a readable tool error rather than a protocol error.
| Tool | Scope | What it does |
|---|---|---|
list_connected_accounts | accounts:read | List the social accounts connected to the workspace, including platform, handle, follower count and connection health. |
list_posts | posts:read | List the posts in the workspace with optional lifecycle status filtering. Drafts, scheduled posts, published posts and failed posts are all returned. |
create_post | posts:write | Create a draft post for one or more connected accounts. Supply a scheduledFor timestamp to schedule it straight away, otherwise the post is saved as a draft. |
update_post | posts:write | Update the content of a post that has not been published yet. Editing a team post creates a new version that has to be approved again. |
schedule_post | calendar:write | Set the publication time of a post. The timestamp must be in the future. Team workspaces with an approval policy require the post version to be approved first. |
publish_post | publish:write | Publish an approved or scheduled post. Use preflight to check that every target is ready before publishing, and execute_now to publish immediately. |
get_post_analytics | analytics:read | Read the performance of a published post, including impressions, reach, engagement and the latest snapshot per platform. |
list_calendar_posts | calendar:read | List the posts scheduled inside a date range. Both bounds are inclusive and are given as ISO 8601 timestamps. |
submit_post_for_approval | approvals:write | Send a draft post into the team approval workflow. Needs an API key created in the team workspace the post belongs to. |
list_pending_approvals | approvals:read | List the posts that are waiting for approval in the API key's team workspace. |
get_account_health | accounts:read | Read the connection health of the connected social accounts. Accounts that need to be reconnected are reported with a reason. |
get_usage_and_quota | billing:read | Read the workspace's plan and its monthly usage (X posts, X posts with links, AI credits) against the plan's allowances. |
Scopes
A key with no explicit scope receives the read only defaults. Write access is always an explicit choice.
posts:readRead posts and drafts
posts:writeCreate and edit posts
publish:writePublish posts
accounts:readRead connected accounts
accounts:writeManage connected accounts
analytics:readRead analytics
calendar:readRead the content calendar
calendar:writeWrite to the content calendar
approvals:readRead the approval inbox
approvals:writeSubmit posts for approval
billing:readRead plan and quota usage
media:readRead the media library
media:writeUpload and edit media
automation:readRead queues, ideas, feeds and webhooks
automation:writeManage queues, ideas, feeds and webhooks
Safety model
The key is the only credential an agent needs, and it is narrow on purpose.
Scoped by default
Every tool declares the scope it needs. A call that the key is not allowed to make is refused before it reaches your data.
Bound to one tenant
A key belongs to one user and optionally to one team. Records that belong to another workspace are reported as missing, so ids cannot be probed.
Rate limited
Each key may make 120 calls per minute. The limit is reported back with a reset timestamp when it is reached.