The write tools here order real work — articles, citations, engagement — and each one costs money. A single confident tool call from a model could commission a campaign. Requiring a second confirm turns that into a decision rather than an accident, and it is the reason the write half is usable at all. On the read side, having citations as first-class objects rather than just mention counts is what makes the data actionable: you can see which pages the assistants actually pointed at.
CrowdReply's endpoint. Its own page names 23 tools, 14 read and 9 write, split across visibility and citations, brands and projects, Reddit listening, and tasks and billing.
- get_visibility_overview and get_visibility_trend — how visible a brand is in LLM answers, and which way it is moving
- list_llm_mentions and list_llm_citations — the individual mentions and the citations behind them
- list_brands, get_brand, list_projects, list_prompts, list_tracked_keywords — what is being monitored
- find_reddit_threads and list_mentions — listening across Reddit and elsewhere
- create_task, create_group_task, cancel_task, refund_task — order citations, articles or engagement
- get_balance and get_usage_metrics — credits and consumption
OAuth 2.0 with PKCE and dynamic registration, scopes read and write, or an x-api-key header. Credit-spending writes go through a two-step confirmation.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen