Sign up for a Hive Moderation account, create a project, and obtain your API key from the project settings
POST the raw chat message text to the Hive Text Moderation synchronous endpoint with the Authorization: Token YOUR_API_KEY header and Content-Type: application/json
Parse the response output array for class labels and associated scores; Hive returns confidence scores per moderation class (hate speech, bullying, profanity, threats, etc.)
Apply a confidence threshold (commonly 0.5 or higher) per class to decide whether to block, flag for review, or pass the message through
Log flagged messages with their Hive response payload for human reviewer audit trails and model feedback
Known gotchas
The default maximum text input length is 1024 characters per call; chat messages exceeding this should be truncated or split, though most in-game chat messages fall within the limit
Hive returns scores per class, not a single overall score; you must define per-class thresholds aligned to your community standards rather than applying one universal cutoff
Synchronous endpoints are optimized for low-latency use but rate limits apply; implement exponential backoff and queue overflow handling to avoid dropped messages during traffic spikes
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp