AT Protocol MCP Server

Appearance

discover_communities ​

Discover clusters of users who interact around a topic. Searches recent posts for the topic, groups the authors by reply interactions, and returns each cluster that meets the minimum size.

Authentication ​

Enhanced - This tool works without authentication but returns richer results when authenticated.

Parameters ​

ParameterTypeRequiredDefaultDescription
topicstringYes-Topic to discover communities around (used as the post search query).
maxResultsnumberNo20Maximum number of communities to return. Must be between 1 and 50.
minCommunitySizenumberNo5Minimum number of members a cluster must have to be returned. Must be 2 or greater.
includeMetricsbooleanNotrueWhether to include the metrics object on each community.

How It Works ​

  1. Searches recent posts matching topic (app.bsky.feed.searchPosts, up to 100 posts).
  2. Groups posts by author, tracking each author's post count and engagement.
  3. Reads reply parents to build interaction edges between authors.
  4. Clusters authors that reply to one another, keeping clusters that meet minCommunitySize.

Derived labels

A community's name is a generated label ("<topic> Community N"), and its description is a generated sentence built from real cluster data (member count and the top contributors' handles). They are summaries of the discovered cluster, not group names that exist on Bluesky.

Response ​

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

typescript
{
  success: boolean;
  communities: Array<{
    name: string;     // generated label, e.g. "web development Community 1"
    topic: string;
    size: number;     // number of members in the cluster
    coreMembers: Array<{
      did: string;
      handle: string;
      displayName?: string;
      avatar?: string;
      followersCount: number;  // from the search result's author profile
      postsCount: number;      // posts about the topic in this sample
      relevanceScore: number;  // engagement + postCount-weighted
    }>;
    activityLevel: 'high' | 'medium' | 'low';  // from avg posts per member
    description: string;  // generated summary sentence
    metrics?: {
      avgFollowerCount: number;
      totalPosts: number;
      interconnectedness: number;  // reply interactions / member count
    };
  }>;
  insights: string[];
}

Examples ​

Discover Tech Communities ​

json
{
  "topic": "web development",
  "maxResults": 20,
  "minCommunitySize": 5
}

Larger Clusters Only ​

json
{
  "topic": "climate change",
  "maxResults": 10,
  "minCommunitySize": 10,
  "includeMetrics": true
}

Quick Discovery (No Metrics) ​

json
{
  "topic": "photography",
  "maxResults": 15,
  "includeMetrics": false
}

Error Handling ​

Common errors:

  • InvalidRequest: Invalid topic or parameters
  • RateLimitExceeded: Too many requests in a short period

When no clusters meet the criteria, communities is empty and insights suggests trying a more popular topic or lowering minCommunitySize.

Community Metrics ​

  • Size: Number of members in the cluster.
  • Activity Level: Derived from average posts per member within the sample.
    • High: average >= 3 posts per member
    • Medium: average between 1.5 and 3
    • Low: average < 1.5
  • Interconnectedness: Reply interactions divided by member count. A larger value means members reply to one another more within the sample. (This is a ratio, not a 0-100 percentage.)
  • Relevance Score: Per-member ranking from topic engagement plus post count; used to order coreMembers.

Core Members ​

coreMembers lists the top members of a cluster (up to 10), ranked by relevance score. Membership and engagement are derived from the sampled topic posts, so they reflect this snapshot rather than long-term community activity.

Rate Limiting ​

Calls are rate limited per tool: 100 requests per minute per tool. This tool performs one post search per request.

See Also ​