← UpdatesJune 14, 2026
Feature

Bloomberry MCP Server: AI-Writing Guardrails for Employee Advocacy

Bloomberry's MCP server lets Claude and Cursor scan employee advocacy drafts for AI-sounding language, banned phrases, and sentence-DNA patterns before employees publish.

Feature A
Feature B
Feature C
Feature D

Why we built this

Employee advocacy content fails when it sounds generic, synthetic, or off-brand. AI detectors try to prove authorship — that is the wrong problem. The real business problem is quality: stopping AI-sounding language from reaching the feed before employees publish.

How it works

Bloomberry now exposes AI-writing guardrails through a real MCP server using stdio transport. V1 checks drafts against 5,698 production AI-writing signal entries — backed by Bloomberry's AI Sentence DNA research corpus of 7,622 audited entries (publicly labeled 7,400+). It is not an AI detector and does not claim to prove authorship. Three callable tools: scan_ai_sounding_text, check_banned_phrases, and analyze_sentence_dna.

What's next

Authenticated hosted MCP/API, Voice Memory comparison to detect voice drift, team-specific guardrails, and approval workflow integration.

← Previous
Bloomberry Market Intelligence OS