AT Protocol MCP Server

Appearance

monitor_keywords ​

Monitor the firehose for posts containing specific keywords. Returns matching posts from the event buffer.

Authentication ​

Optional - This tool works without authentication but requires the firehose to be started first using start_streaming.

Parameters ​

ParameterTypeRequiredDefaultDescription
keywordsstring[]Yes-Array of keywords to monitor. At least one keyword is required.
limitnumberNo20Maximum number of matching posts to return. Must be between 1 and 100.
caseSensitivebooleanNofalseWhether keyword matching should be case-sensitive.

Response ​

typescript
{
  success: boolean;
  keywords: string[];
  matches: Array<{
    keyword: string;
    post: {
      uri: string;
      text: string;
      author: string;
      createdAt: string;
    };
    seq: number;
    receivedAt: string;
  }>;
  totalMatches: number;
  totalScanned: number;
}

Examples ​

Monitor Single Keyword ​

json
{
  "keywords": ["bluesky"],
  "limit": 10
}

Monitor Multiple Keywords (Case-Sensitive) ​

json
{
  "keywords": ["AI", "ML", "LLM"],
  "limit": 50,
  "caseSensitive": true
}

Monitor Hashtags ​

json
{
  "keywords": ["#technology", "#programming", "#webdev"],
  "limit": 25
}

Error Handling ​

Common errors:

  • InvalidRequest: No keywords provided or invalid parameters
  • StreamingNotStarted: Firehose streaming must be started first using start_streaming
  • RateLimitExceeded: Too many requests in a short period

Best Practices ​

  1. Start Streaming First: Always call start_streaming before using this tool
  2. Use Specific Keywords: More specific keywords reduce noise and improve relevance
  3. Adjust Limit: Start with a smaller limit and increase as needed
  4. Case Sensitivity: Use case-sensitive matching for acronyms and proper nouns
  5. Buffer Size: The tool scans the event buffer, which has a maximum size - older events may be dropped
  6. Combine Keywords: Use multiple related keywords to capture variations
  7. Monitor Hashtags: Include hashtags to track specific topics or trends

Rate Limiting ​

This tool reads from the local event buffer and is not subject to AT Protocol API rate limits. However, the firehose connection itself has limits:

  • Maximum 1 concurrent firehose connection per client
  • Event buffer size is limited (typically 1000 events)
  • Older events are dropped as new ones arrive

See Also ​