MCP Server
Connect Claude Code, Claude Desktop, Cursor, OpenClaw or any MCP-compatible client directly to HotMention. Your agent can list the hot leads, read the draft and the posting guidance, regenerate the draft with instructions, record the reply it posted from your account, watch the Reddit budget and read the Citation Share scoreboard. Available on all plans, including Free. Server version 2.0.0.
Endpoint
Stateless streamable HTTP (JSON-RPC 2.0). No session or SSE stream required.
https://hotmention.com/api/mcp
Authenticate with a Bearer token, generated in Settings → API. Every key is bound to one project; pass project_id on any tool to act on another project the key's owner can access.
Authorization: Bearer hm_your_key
Claude Code
Add the server with a single command:
claude mcp add --transport http hotmention https://hotmention.com/api/mcp \ --header "Authorization: Bearer hm_your_key"
Cursor / Claude Desktop / OpenClaw
Add this to your MCP servers config (e.g. mcp.json or claude_desktop_config.json):
{
"mcpServers": {
"hotmention": {
"url": "https://hotmention.com/api/mcp",
"headers": { "Authorization": "Bearer hm_your_key" }
}
}
}Available Tools
Every tool accepts an optional project_id. Results are the same JSON shapes as the REST API; errors come back as tool errors with the same codes (for example budget_exhausted).
| Tool | Description | Args |
|---|---|---|
| The reply loop | ||
list_hot_leads | HOT leads waiting for a reply. Each carries the draft, the thread URL, the subreddit survival score, posting guidance (ok / careful / avoid) and the current Reddit reply budget. Respect budget.exhausted and postingGuidance.level = 'avoid'. | platform, since, limit, include_ai_cited |
get_lead | The full card for one lead: post text, why it scored, draft, reply status and guidance rules. | id |
regenerate_draft | Write a new draft reply, optionally steered by instructions such as 'mention the free tier, no link'. Counts against the draft quota on limited plans. | id, instructions |
mark_replied | Record that you posted a reply from your own account. Pass reply_url (the comment permalink) so it can be verified on day 1, 3 and 7. Fails with budget_exhausted when the Reddit budget is used up unless force is true. Idempotent. | id, reply_url, account, force |
skip_lead | Skip a lead with an optional reason. | id, reason |
get_reply_budget | The project's Reddit reply budget: caps, used today and this week, remaining, and the suggested time for the next reply (60-minute spacing). | — |
| Citation Share | ||
get_visibility | The scoreboard: cited in N of M checks across ChatGPT, Google AI Overviews, Perplexity and Claude, the delta vs the previous run, per-engine tiles, and the prompts where competitors are cited instead, with the source feeding each answer. | — |
| Mentions and keywords | ||
search_mentions | Search scored mentions for the project. Returns posts with intent scores, verdicts and draft replies (HOT, WARM and LOW). | keywords, platform, score_min, since, limit |
list_keywords | List tracked keywords with active / auto-discovered flags. | — |
add_keyword | Add a keyword to track. Respects the plan's keyword limit. | term |
get_usage | The account's plan, lead usage for the current billing period, and limits. | — |
Try it
Prompts that work well:
- “List my hot leads from the last 24 hours and tell me which three are worth answering and why.”
- “Regenerate the draft for lead_8f2k: shorter, mention the free tier, no link. Then I'll post it myself.”
- “I posted it: https://www.reddit.com/r/.../comments/.../ — mark lead_8f2k replied.”
- “How is my Citation Share this week, and which prompts do I lose to competitors?”
Want the agent to post as well? It can, from your own Reddit account, with the tools above as its guardrails. See the Autopilot guide and the recipes.