OmniSocials Agent Skills

AI agent skills for managing social media across 13 platforms (14 channels) via OmniSocials. Give your AI agent the ability to create, schedule, and publish posts directly from your IDE or terminal.

Works with Claude Code, Cursor, Windsurf, GitHub Copilot, Codex, Gemini CLI, and many others.

Supported Platforms

Instagram, Facebook, LinkedIn (personal profile + company page), YouTube, TikTok, X (Twitter), Pinterest, Bluesky, Threads, Mastodon, Google Business, Telegram (channels and groups), Discord (channels)

Setup

1. Install the skill

npx skills add OmniSocials/omnisocials-agent-skills

This installs the skill (including the CLI at <skill-path>/scripts/omnisocials.js) into your agent's skills directory. Alternatively, clone the repo and run the CLI from skills/omnisocials/scripts/omnisocials.js (the old scripts/omnisocials.js path still works as a shim).

2. Copy your API Key

Get your API key from Settings > API in the OmniSocials app.

3. Run the setup command

Ask your agent "Help me set up the OmniSocials skill", or run the setup yourself from the install location:

./skills/omnisocials/scripts/omnisocials.js setup

4. Start using it

Ask your AI agent things like:

• "Show my connected social accounts"
• "Create a post for Instagram and LinkedIn with this image"
• "Schedule a reel for TikTok tomorrow at 9am"
• "Show my scheduled posts"
• "How are my posts performing this week?"
• "Upload this image and create a Pinterest pin"

Commands

70 commands covering the full v1 API:

Features

• 13 platforms (14 channels) from one tool
• Posts, Stories, and Reels with platform-specific options
• Platform-specific controls: Pinterest boards and product tags, YouTube metadata, TikTok privacy, Instagram covers, X reply settings
• Per-platform media: Different images/videos for different platforms in the same post
• Image fit per channel (unreleased): a feed image outside what a channel shows uncropped is cropped or kept per channel (--media-fit instagram=keep,pinterest=crop); media:check shows how each channel would show it first
• Alt text: Per-media accessibility descriptions, delivered to Mastodon, Bluesky, X, Pinterest, Instagram (images) and LinkedIn (images)
• Analytics: Post-level, bulk, workspace overview, account-level metrics, and best posting times
• Post tags: label posts by hook or format (--tags, tags:apply for older posts in bulk) and compare the tags against your normal, fair across channels (analytics:tags)
• Social inbox: Read and reply to DMs, comments, and mentions
• Comment to DM: answer a keyword in an Instagram or Facebook Page comment with a DM, once per person, with a public reply under the comment; on a new post (--dm-keywords, --dm-message) or an existing one (automations:create), with the activity per person (automations:events). On X (--platform x) an automation answers a keyword in a DM to the account
• Approvals: Send a post through an approval workflow, approve or reject it, and read the review (who decided, the reason, the comments)
• X statistics on credits: read the credit balance and each X account's estimate (credits:get), choose how often an X account's posts are checked and its followers counted (accounts:x-statistics), and check one X post now (analytics:refresh)
• Webhooks: Get notified when posts are scheduled, published, fail, or are approved or rejected in an approval workflow, and when X statistics pause because credits ran out
• Limits and feature requests: see what OmniSocials cannot do, why, and what to do instead (limits), ask the OmniSocials team for it (feature:request), and follow the answer (feature:list, feature:get)
• Zero dependencies: Uses Node.js 18+ built-in fetch

Alternative: MCP Server

For agents that call tools instead of shell commands, add the hosted OmniSocials MCP server. It signs in with OAuth (no API key) and has every tool. In Claude Code:

claude mcp add --transport http --scope user omnisocials https://mcp.omnisocials.com
claude mcp login omnisocials

Setup for other agents: omnisocials.com/mcp. The @omnisocials/mcp-server npm package runs a local copy with an API key, with fewer tools than the hosted server.

Claude Code and Grok Build plugin

This repo is also an installable plugin marketplace for Claude Code and Grok Build. The plugin bundles the skill plus the hosted OmniSocials MCP server.

• Claude Code: /plugin marketplace add OmniSocials/omnisocials-agent-skills, then /plugin install omnisocials@omnisocials-skills.
• Grok Build: grok plugin install OmniSocials/omnisocials-agent-skills --trust, or grok plugin marketplace add OmniSocials/omnisocials-agent-skills and install it from /plugins. It is not in xAI's official marketplace.

On first use, sign in to OmniSocials and approve the workspaces the agent can access.

The MCP connection uses OAuth. No API key is needed for it.

Cursor

Install this repository as a Cursor plugin. Open Customize, import OmniSocials/omnisocials-agent-skills as a marketplace, and install omnisocials. To try it locally, copy the repo to ~/.cursor/plugins/local/omnisocials and reload Cursor. Cursor loads the skill from skills/ and the hosted MCP server from mcp.json.

The MCP server signs in with OAuth (scopes social:read and social:write). No API key is stored for it. The CLI still uses OMNISOCIALS_API_KEY.

Network endpoints and credentials

For transparency and security review, this plugin only calls OmniSocials services:

• https://api.omnisocials.com/v1 (skill CLI; authenticated with the OmniSocials API key you provide during setup, stored in your local config)
• https://mcp.omnisocials.com (hosted MCP server; OAuth 2.1 sign-in, no key stored by the plugin)

The plugin sends no telemetry and reads no credentials other than its own API key.

Links

• OmniSocials
• API Documentation
• Integrations Guide
• MCP Server
• llms.txt (docs index for AI agents)

License

MIT