FAQ
AI social media agent FAQ
What is the ViralHunt MCP server and what can an agent do with it?
viralhunt-mcp on npm exposes the ViralHunt API as 51 typed tools over the Model Context Protocol, so Claude Desktop, Claude Code, Cursor, Windsurf or any MCP client can read rising posts across 13 sources, the top hashtags, trending sounds and the best time to post, fill your brand templates, leave complete posts in Drafts, review them against your content policy and, once a person approves, schedule them to 11 networks.
Will an AI agent publish to my social accounts without my approval?
No. A member or an agent token can only leave a post in Drafts or send it to Review; only the organization owner, an admin, or a collaborator the owner marked as trusted can approve, and an OK a writer gives their own post never publishes it. A token can also be minted with fewer rights (reader, writer, reviewer) so the agent cannot reach the publish actions at all.
Which AI agents and automation tools work with ViralHunt?
Claude on claude.ai through the skill; Claude Desktop, Claude Code, Cursor and any MCP client through viralhunt-mcp; ChatGPT through a Custom GPT with one Action; OpenClaw through its skill hub; n8n and Make through HTTP nodes; your own agent through the REST API with a personal token, or by webhook from the composer. Every one of them uses the same key.
How much does API and MCP access cost?
The Free plan ($0, no card) includes a key: 24 content queries a day at 60 requests an hour, on posts 72 hours old and older, with the engagement numbers withheld. Creator ($29/mo, 1 brand, up to 11 connected accounts) lifts the daily cap, shows posts as they break with their numbers, and adds publishing at 100 requests an hour. Studio ($99/mo, 4 brands) has 2,000 requests an hour and Agency ($199/mo, 10 brands) 5,000.
Can the agent run on a schedule, like a social media manager would?
Yes. A recipe is a standing order for a brand: what to post from (the radar, the quote base, the calendar of dates), on which template, to which networks, how often and at what time. The agent reads it through the API, runs it on its own clock, leaves the posts as drafts and stamps each run. The skill ships the loop as a command, so one line starts the daily job on a page or an account.
Can I connect an agent I already run instead of a token?
Yes. Connect it by webhook from the Agents page: ViralHunt signs a POST to your URL when someone asks it something from the composer, and your agent answers once. An agent with no public URL polls the same tasks through the API. OfficeMatrix, the open source office of agents, has the connection built in.