Documentation

Slop Sentry User Guide

Learn how to connect Slop Sentry to your workflow so your AI-assisted drafts go out free of AI tells, with concrete claims and clean AP style.

Connect Your AI Agent

Choose your preferred AI solution to see the easy setup instructions.

⚠️

API Key Required

You must replace YOUR-API-KEY in the snippets below with your actual Slop Sentry token. You can find your token in your Account Dashboard.

Setup for Automatic (AI Agent)

Get API Key: Sign up for a plan to receive your private access token.
Copy Prompt: Copy the prompt on the right and paste it into your AI assistant.
Replace Key: Make sure to replace YOUR-API-KEY with your actual token before sending.

I want to add a new MCP server. Please update my configuration to include:
- Name: slop-sentry
- URL: https://mcp.slopsentry.ai
- Authorization: Bearer YOUR-API-KEY

Works best with Claude Code, Cursor, or other agents with file access.

Setup for Claude Code

Get API Key: Sign up for a plan to receive your private access token.
Add Server: Run the command below in your terminal to register the server. Include --transport http so Claude treats the URL as an HTTP MCP server instead of a local stdio command, and --scope user so Slop Sentry is available in every project (not only the folder where you ran the command).
Verify: Type `claude mcp list` to ensure Slop Sentry is active.

claude mcp add --scope user --transport http slop-sentry https://mcp.slopsentry.ai --header "Authorization: Bearer YOUR-API-KEY"

Without --transport http, Claude Code misreads the URL as a stdio command. Without --scope user, the default local scope only registers the server for the current project — you would need to re-add it in every repo. Use the latest Claude Code.

Setup for ChatGPT

Use the desktop app: These steps are for the ChatGPT macOS/Windows desktop app, not chatgpt.com in a browser. Mobile apps cannot add MCP servers — configure on desktop, then use the server on other supported surfaces.
Get API Key: Sign up for a plan to receive your private access token from your Account Dashboard.
Open MCP settings: In the desktop app, open Plugins, then Settings → Plugins → MCPs (labels may also appear as Settings → MCP servers). Select Add Server — not “Create App”.
Choose Streamable HTTP: The Add Server dialog defaults to STDIO (local command). Switch the transport to Streamable HTTP. Leave STDIO selected only if you intend to run a local executable — that is not how Slop Sentry works.
Fill in the form: Name: slop-sentry (or similar)
URL: https://mcp.slopsentry.ai
Authenticate: ChatGPT has no plain “paste your Bearer token” field. Use Headers instead:

Header name: Authorization
Header value: Bearer YOUR-API-KEY
(Include the word Bearer, then a space, then your real API key in place of YOUR-API-KEY.)

Bearer token env var: Leave this empty when using Headers. That field is only for an operating-system environment variable name (for example SLOPSENTRY_TOKEN) — not the token itself. Use it only if that variable is already set in the environment that launches ChatGPT.
Save and verify: Save the server, then Restart when prompted. In a chat, type /mcp — Slop Sentry should appear as enabled with tools discovered (for example slop_sentry). If the server lists but tools fail, re-open the entry and confirm the Authorization header is present and correct.

Transport: Streamable HTTP (not STDIO)
Name: slop-sentry
URL: https://mcp.slopsentry.ai

Headers
  Authorization: Bearer YOUR-API-KEY

# Alternative (advanced): Bearer token env var = SLOPSENTRY_TOKEN
# (OS env name only — export the token before launching ChatGPT)

Desktop app required for this MCP UI. ChatGPT Plus, Pro, or Enterprise recommended. Web (chatgpt.com) uses a different Plugins / connectors path and may not match these steps. Config is shared with Codex on the same machine.

Setup for Codex

Get API Key: Sign up for a plan to receive your private access token.
MCP Settings: Open the Codex MCP settings panel or edit `~/.codex/config.toml`.
Add Server: Paste the configuration into your server list.

[mcp_servers.slop-sentry]
url = "https://mcp.slopsentry.ai"
headers = { Authorization = "Bearer YOUR-API-KEY" }

Restart the Codex agent after saving your changes.

Setup for Claude Cowork

Get API Key: Sign up for a plan to receive your private access token.
Connectors: In the Cowork dashboard, go to Workspace Settings > Connectors.
Add Custom: Select "Add Custom MCP" and fill in the details below.

URL: https://mcp.slopsentry.ai
Header: Authorization: Bearer YOUR-API-KEY

Requires Workspace Admin permissions to add new connectors.

Setup for Suits.ai

Get API Key: Sign up for a plan to receive your private access token.
Open Integrations: In your Suits.ai console, navigate to Settings > Integrations and create a new MCP connection.
Enter Details: Use Slop Sentry as the title and enter the MCP endpoint URL shown on the right. You can leave Description blank.
Add API Key: Expand Advanced. Under HTTP headers, enter Authorization for Name and Bearer YOUR-API-KEY for Value, replacing YOUR-API-KEY with your token.
Authorize: Click Authorize to finish connecting Slop Sentry.

Title: Slop Sentry
MCP endpoint URL: https://mcp.slopsentry.ai

Advanced > HTTP headers
Name: Authorization
Value: Bearer YOUR-API-KEY

Operating Modes

Two Ways to Protect Your Writing

Slop Sentry offers two workflows depending on whether you want a quick fix or full manual control. Once the MCP server is connected, paste prompts like these into Claude, ChatGPT, or your coding agent — include your draft in the same message.

Recommended

Option A: The Sentry

The slop_sentry tool is the fastest route to a finished piece. It runs a four-phase pipeline (Substance, Hokiness, AI Tells, and AP Style) and rewrites your prose until every check reports CLEAN.

Automatic fixes across all four checks in one call
Re-runs the checks until clean, never re-processing the same text twice
A separate AI reviewer reads your draft fresh and judges whether it says anything concrete to your target audience
Clean prose comes back in seconds

Example prompts

Draft a press release announcing our Series B. Then run slop_sentry with content_type: press_release and show me only the cleaned final version.

Generate, then gate — one call before anything ships.

Here's my customer email below. Run slop_sentry with content_type: email. Keep my facts, names, and CTA; strip AI-sounding padding.

Proof an existing draft without losing specifics.

Run slop_sentry on this LinkedIn post with content_type: linkedin. Preserve the opening hook and the three metrics.

Name the content type so channel rules apply.

Use slop_sentry on this blog draft, but set skip_ap_style and skip_hokiness. I only want substance and AI-tell fixes.

Skip phases when you want a lighter pass.

Option B: Manual Pipeline

For maximum control, use the individual tools one at a time — audit first, rewrite yourself, then re-check until clean.

1. Detection: Use quality_check to run all four checks and get a unified violation report.
2. Targeted Checks: Run substance_check, hokiness_check, or ap_style_check individually for focused audits — or use guard_sentry for a Guard-only automated rewrite loop.
3. Re-Check: Run quality_check again with recheck: true to verify your manual edits.

Example prompts

Run quality_check on this article with content_type: article. Don't rewrite yet — list every finding and why it failed.

Step 1 — full audit before you touch the draft.

Run substance_check on the opening three paragraphs. Tell me what's vague or empty so I can rewrite them myself.

Step 2 — focused judgment when you want to keep the pen.

Run guard_sentry on this blog post to strip AI tells only. Show me a short before/after of what changed.

Step 2 alt — Guard-only rewrite without Substance, Hokiness, or AP Style.

I rewrote the flagged sections. Run quality_check again with recheck: true and the same session_id. If anything remains, suggest one fix per finding — don't rewrite the whole piece.

Step 3 — verify edits in the same session until CLEAN.

Day-to-day writing management

Treat Slop Sentry as a gate in your normal drafting loop. Paste the draft (or ask the agent to draft first), name the content type, then choose Option A for a clean pass or Option B when you want to review findings yourself.

Before you publish

"Before you finalize, run slop_sentry on the full draft with the right content type and return the cleaned text."

Iterate on one section

"Only check this revised paragraph with quality_check and recheck: true. Leave the rest of the doc alone."

Compare two versions

"Run slop_sentry on Version A and Version B separately, then tell me which reads cleaner and why — don't merge them."

Standing instruction

"For every prose draft in this chat, call slop_sentry before showing me the final text. Ask me for the content type if it's unclear."

Content Types

Match the Check to the Channel

Name a content type in your prompt so rules match the format — LinkedIn paragraph limits, press-release tone, email directness, and more.

"Run slop_sentry on this post with content_type: linkedin."

Enforces mobile-friendly paragraph limits and caps repetitive openers.

Press release

"Run slop_sentry on this release with content_type: press_release."

Keeps AP Style and newsroom tone in the pipeline for external announcements.

Email / proposal

"Run slop_sentry with content_type: email (or proposal) on the text below."

Favors directness and strips formal AI padding common in outbound copy.

Supported Content Types

Tailor the Sentry's behavior by naming the content type, from press releases and LinkedIn posts to proposals and email:

default article blog linkedin email proposal guide pptx image_text press_release

No MCP? No Problem.

Using SKILL.md as a System Prompt

If your AI assistant doesn't support MCP yet, or you want it to carry Slop Sentry's full editing rulebook rather than just call the tools, use our SKILL.md file as its system instructions.

Copy its contents from your dashboard and paste them into your "System Instructions" field. The AI gets the full logic of Slop Sentry, even without the automated tools.

Developer Tools

The Slop Sentry CLI

For automated workflows (scripts, publishing pipelines, and developer setups) we provide a Python command-line tool that runs the same checks.

Terminal · slopsentry

$ slopsentry check draft.md --type blog
❌ FAILED - 2 violations.
HYPOPHORA (1): Self-asked Q&A...
BUZZWORD (1): "delve"
$ slopsentry rewrite draft.md --provider remote
Job initiated. ID: 4f2a... Polling...
Content is now CLEAN.