AT Protocol MCP Server

Appearance

analyze_moderation_status ​

Analyze moderation status of a post or user. Returns content labels, moderation decisions, and personal moderation state (blocks, mutes). Subject can be a DID (for users) or AT-URI (for posts).

Authentication ​

Required: Yes (Private tool). Personal moderation state (your blocks and mutes of the subject) is only meaningful when authenticated.

Parameters ​

ParameterTypeRequiredDefaultDescription
subjectstringYes-The thing to analyze. For a user, pass a handle (alice.bsky.social) or DID (did:plc:...). For a post, pass its AT-URI (at://...). The subject type is inferred from whether the value starts with at://.
includeLabelsbooleanNotrueWhether to include content labels in the response.

Response ​

Tool results are returned as stringified JSON text. The illustrative shape is:

typescript
{
  success: boolean;
  subject: string;        // echoes the subject you passed in
  subjectType: 'user' | 'post';
  moderation: {
    labels?: Array<{
      src: string;
      uri: string;
      val: string;
      cts: string;
    }>;
    blocked?: boolean | string; // boolean for user subjects; for post subjects this is the author's block-record AT-URI (string) or undefined
    muted?: boolean;          // you mute this user
    blockedBy?: boolean;      // this user blocks you
    blocking?: string;        // AT-URI of your block record, if any
    mutedByList?: boolean;    // muted via a list (users only)
    blockingByList?: boolean; // blocked via one of your lists (users only)
  };
  analysis: {
    hasContentWarnings: boolean;
    isNSFW: boolean;
    isSpam: boolean;
    requiresWarning: boolean;
    safetyLevel: 'safe' | 'warning' | 'restricted' | 'blocked';
  };
}

For a post subject, only blocked, muted, and blockedBy (reflecting the post author's relationship to you) are populated; the list-based fields apply to user subjects only. Note that for a post subject blocked carries the post author's block-record AT-URI string (or undefined) rather than a boolean, unlike the user path, which coerces it to a boolean via !!.

Examples ​

Analyze User Moderation Status ​

json
{
  "subject": "did:plc:abc123",
  "includeLabels": true
}

Analyze Post Moderation Status ​

json
{
  "subject": "at://did:plc:abc123/app.bsky.feed.post/xyz1",
  "includeLabels": true
}

Quick Moderation Check (No Labels) ​

json
{
  "subject": "alice.bsky.social",
  "includeLabels": false
}

Error Handling ​

A missing or unresolvable subject (handle/DID/post not found, or a post that is itself blocked) surfaces as the underlying AT Protocol error. Per-tool rate limiting also applies (see below).

Subject Types ​

  • User - pass a handle (alice.bsky.social) or DID (did:plc:abc123). Returns profile-level labels and your blocks/mutes of that user.
  • Post - pass an AT-URI (at://did:plc:abc123/app.bsky.feed.post/xyz1). Returns post-level labels and the author's moderation state relative to you.

Content Labels ​

Common label values:

  • porn: Adult sexual content
  • sexual: Sexually suggestive content
  • nudity: Nudity (artistic or otherwise)
  • graphic-media: Graphic violence or disturbing imagery
  • spam: Spam or misleading content
  • impersonation: Impersonation of another user

Safety Levels ​

  • safe: No content warnings, safe for all audiences
  • warning: Has content warnings, user discretion advised
  • restricted: Restricted content, may be hidden by default
  • blocked: Blocked content, should not be displayed

Moderation State ​

Personal Moderation ​

  • blocked: You have blocked this user
  • muted: You have muted this user
  • blockedBy: This user has blocked you
  • blocking: AT-URI of your block record for this user, if any
  • mutedByList: Muted via a list
  • blockingByList: Blocked via one of your lists

Public Moderation ​

  • labels: Content labels applied by moderators or automated systems
  • hasContentWarnings: Whether content has any warnings
  • isNSFW: Whether content is marked as NSFW
  • isSpam: Whether content is marked as spam

Rate Limiting ​

Calls are rate limited per tool to 100 requests per minute (see the error handling guide).

See Also ​