Skip to content

Prose Coach MCP

Remote MCP server (Streamable HTTP, no auth) for Prose Coach. A deterministic writing filter that marks 43 patterns that make text read as AI-generated — banned vocabulary and structural tells (triplets, negation formulas, summary closers, colon chains, flat rhythm) — with every finding quoting the line that triggered it and attaching the fix.

Server type: Remote (Streamable HTTP)
Auth: None
Endpoint: https://www.prose.coach/mcp
Pricing: Free tier (3 drafts/day, 12,000 chars per request); PRO $5/mo or $50/yr
Category: Content / Writing

Why This Matters for Operators

Every AI assistant reaches for the same moves across every draft. Prose Coach is the first MCP purpose-built to de-AI drafts before they ship: a fixed rule set runs over your text, flags what's there, and density is scored against an accumulation threshold — one marker is noise, six is a habit. Because no language model reads your text, the same input always returns the same findings — unlike probabilistic "99% AI-written" detectors whose scores can't be checked against anything.

Tools & Capabilities

Tool Tier What it returns
scan_draft Free Every finding with the line quoted and the fix attached
coach_draft PRO Scan + the written rule behind each finding, scope map for content type, judgment layers
verify_revision PRO Before/after diff: patterns resolved, reduced, still present, or newly introduced

Two prompts (scan, polish) ship alongside the tools. Re-scanning a draft after edits doesn't count against the free daily limit.

Installation

claude mcp add prose-coach --transport http https://www.prose.coach/mcp

Per-client walkthroughs (Claude Code, Codex, Cursor, VS Code): prose.coach/connect

Configuration

{
  "mcpServers": {
    "prose-coach": {
      "type": "http",
      "url": "https://www.prose.coach/mcp"
    }
  }
}

No API key. A browser scanner at prose.coach runs the same engine if you want to try it before connecting anything.

Business Relevance

  • Marketing teams can run every email, ad, and social post through scan_draft before it ships — structural tells (triplets, summary closers) are exactly what makes mass-produced AI copy feel uncanny
  • Content operators get a free tier covering 3 drafts/day — enough for daily posting cadences without a subscription
  • Founders and executives get a deterministic filter instead of an opaque AI-probability score they can't verify
  • Agencies can standardize voice at scale — PRO ($5/mo) gates verify_revision, which diffs the rewrite against the original

Integration with CorpusIQ

Prose Coach pairs with CorpusIQ's business-data connectors on the content side of the stack: an agent pulls live numbers (Shopify orders, GA4 traffic, ad spend) through CorpusIQ, drafts customer-facing copy from that data, then runs the draft through Prose Coach before anything ships. CorpusIQ supplies the data; Prose Coach polishes the prose. The composition is natural for the same workflows covered by CorpusIQ's content-ops tooling: weekly performance recaps, customer win emails, and product update posts — all drafted by agents, all passed through a deterministic writing filter before they reach a human.

Limitations

  • Brand new — listed on mcp.so Aug 12, 2026, no ecosystem track record yet
  • Free tier is a single tool (scan_draft); coach_draft and verify_revision require PRO
  • The 43-pattern rule set targets English-language writing moves
  • It flags patterns and attaches fixes — it does not rewrite; the revision happens in your client
  • Hosted endpoint only (no self-hosted option); no auth means no account-scoped history

See Also