Content moderation for chat and comments: free test keys, message checks, plain-English policies.
Scored across 11 tools
Most tools target distinct resources, but test_policy and test_policy_on_project do nearly the same thing and differ only in how the project is identified, which is easy to misselect. get_player_history also substantially overlaps with list_flagged_messages filtered by player.
All 11 names use consistent snake_case with a verb-first pattern (create_, get_, list_, test_, moderate_). The only minor deviation is the suffixed test_policy_on_project, which breaks the otherwise clean verb_noun shape.
Eleven tools is well within a reasonable range for a moderation API and each maps to a recognizable capability. The only redundancy is the policy-testing pair, which could arguably be one tool.
The surface covers the read and evaluate side of moderation well: projects, settings, usage, flagged messages, player history, moderation, policy/word-rule testing and onboarding. The notable gap is write access: no tool to update thresholds, save a draft policy or word rules, or submit TP/FP reviewer verdicts despite verdicts being displayed.
Creates a working text-moderation API key with no account: 100 messages over 72 hours, standard mode, every category on. Returns api_key and claim_url; opening claim_url keeps the key in a Paxmod account. Limited to 3 keys per address per day.
Inputs: name, source.
The full Paxmod API guide for coding agents: endpoints, request and response shapes, error codes, rate limits and code samples (Node, Python, Unity, Roblox and more).
Inputs: no named parameters.
Returns one player's flagged messages in a project, newest first (up to 50), with a count by reason. Use the player id the game or app sends as user_id. Not available for organizations that keep player data in a separate region unless Paxmod has enabled it.
Inputs: user_id (required), project_id (required), limit.
Returns a project's moderation settings: type, mode, categories and thresholds, the Advanced mode policy, and word rules. The API key is not included.
Inputs: project_id (required).
Returns how many messages a project has sent per day over the last 30 days (up to 90), plus the total for the current calendar month. Counts are updated through the day.
Inputs: project_id (required), days.
Lists a project's most recent flagged messages, newest first, with the reason, confidence, player id, context id and any reviewer verdict (TP or FP). Filter by date, category or player. Up to 50 per call. Not available for organizations that keep player data in a separate region unless Paxmod has enabled it.
Inputs: project_id (required), limit, since, user_id, +1 more.
Lists the signed-in user's Paxmod projects: their own and those in organizations they belong to, with each project's id, name, type (text or image), mode (standard or advanced), organization and the user's role. API keys are not included.
Inputs: no named parameters.
Checks one chat message with Paxmod, using the project's categories and thresholds. Returns flagged or not_flagged with a score per category. Counts as one of the project's messages, like a live request. Needs a Paxmod API key (create_test_api_key gives one).
Inputs: api_key (required), message (required), user_id, context_id.
Runs example messages against an Advanced mode policy and says which would be flagged, with a score and reason for each. Send a draft policy to try changes before saving them; leave it out to test the saved policy. Mark each example with expect "flag" or "pass" and the summary counts how many came back as expected. The project's word rules run first, as they do live. Nothing is stored. No credits are used, except on a free test key from create_test_api_key, where each message checked counts toward its 100 free messages. Up to 50 examples per call.
Inputs: api_key (required), examples (required), policy, threshold.
Runs example messages against one of the user's Advanced mode projects, with a draft policy or the saved one, and says which would be flagged, with a score and reason for each. Mark each example with expect "flag" or "pass" and the summary counts how many came back as expected. The project's word rules run first, as they do live. Nothing is stored and no credits are used. Up to 50 examples per call.
Inputs: examples (required), project_id (required), policy, threshold.
Shows what a project's word rules do to each message: block (flag without checking), bypass (pass without checking), or scan with the text rewritten by remove and replace rules. Send draft rules to try changes before saving them. No model call, nothing stored, no credits used.
Inputs: api_key (required), messages (required), word_rules.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Sentiment, toxicity, entity extraction, PII, translation, summary, QA, fraud scoring, safety audit.
Detect grooming, bullying, fraud, and 16+ online threats across text, voice, image, and video.
Guardrails for LLM output: pass / fail / review verdicts and one recommended action. Hosted or npm.
Pay-per-call profanity/explicit-content detection for AI agents. $0.005 USDC per call, no signup.
Add one secure layer between your agents and this server.
