Use Socifyr with AI agents
Socifyr ships an MCP server with 20 tools. Connect it to Claude, Cursor, ChatGPT, Codex or any MCP client and the agent can list your accounts, publish and schedule posts, retry failures, refresh expired tokens and read analytics, using plain-language prompts. Every tool is a thin wrapper over the same REST API the dashboard uses, so the plan limits and API key guardrails apply to agents too.
How do I connect an agent?
Pick your client. Each page has copy-paste configuration.
| Client | Page | Connection type |
|---|---|---|
| Claude (claude.ai, Claude Desktop, Claude Code) | Use Socifyr from Claude | Remote URL or local server |
| Cursor | Use Socifyr from Cursor | Remote URL |
| ChatGPT | Use Socifyr from ChatGPT | Remote URL with a header |
| Codex | Use Socifyr from Codex | Local server |
| Any other MCP client | MCP setup | Remote URL or local server |
The remote endpoint is https://api.socifyr.com/api/mcp (Streamable HTTP) with the header Authorization: Bearer sk_.... The local server is npx -y @socifyr/mcp with SOCIFYR_API_KEY and SOCIFYR_WORKSPACE_ID set.
What can an agent do with Socifyr?
| Job | Tools |
|---|---|
| Find accounts and their health | list_connections |
| Publish now | post_text, post_image, post_video |
| Schedule | schedule_post, schedule_posts (up to 500 at once) |
| Find and edit posts | list_posts, update_post, publish_draft |
| Check results and recover | check_post_status, retry_post, refresh_connection, reconnect_connection |
| Media | list_media, delete_media |
| Reporting | get_post_analytics, get_follower_growth, compare_accounts, get_usage |
| Writing help | generate_caption |
Full inputs, samples and errors: MCP tools.
What can an agent not do?
Say these plainly to your team before you hand over a key:
- It cannot connect a new social account. Accounts are connected in the dashboard.
- It cannot upload a file from your computer. It can reuse media already in the library or a public URL.
- It cannot create a draft, submit a draft for approval, approve or reject. With a key that requires approval, its posts are held for a human.
- It cannot manage webhooks, API keys, members or workspaces.
- It cannot suggest hashtags. Only captions are available as a tool.
The feature matrix compares every surface.
How do I keep an agent safe?
Give each agent its own API key and set guardrails: a daily post limit, a monthly X spend cap and required approval. A guarded key is deny-by-default on the API, so even a confused agent cannot approve its own posts or lift its own limits. Setup: Approvals and guardrails.
Where do I start?
- Try example agent prompts with the tool calls they trigger.
- Build your own with the content agent walkthrough.
- Compare surfaces in Choose your surface.
FAQ
Does the agent see my API key?
The key is in the client configuration and is sent as a header. Keep it in an environment variable where the client supports interpolation, and rotate it from the dashboard if it leaks.
Which workspace does the agent use?
The key's workspace. Use one key and one connector per workspace. See Multi-workspace setup.