DEV Community: Mohab Abdelkarim

One MCP Server, Three Assistants: Running Walter Across Claude, ChatGPT, and Codex

Mohab Abdelkarim — Mon, 20 Jul 2026 12:46:32 +0000

TL;DR: One MCP server, https://mcp-server.walterwrites.ai/mcp, now works across three completely different AI clients: Claude (stable, native connector support), ChatGPT (beta, via Developer Mode), and Codex (native CLI/IDE support). Same tools, same detection and humanization logic, three different setup flows. This post covers the actual config for all three, what's different about each, and why this is a genuinely interesting case study in MCP interoperability even if you don't care about Walter specifically.

Six months ago, this would've been a one-platform story. Walter's MCP server, which exposes AI text humanization and detection as callable tools, only worked inside Claude. If your team drafted in ChatGPT, or you were building internal tooling on top of Codex, you were stuck with the old copy-paste-detect-copy-paste-humanize loop that MCP exists specifically to kill.

That's no longer true, and the reason it's no longer true is a good, concrete example of what the M×N integration problem MCP was built to solve actually looks like when it's solved. One server. Three clients. Zero custom integration code written on Walter's side to support any of them.

Here's the setup for all three, the actual differences in how each client handles tool calling, and a few things that will save you time if you're setting this up yourself.

Why this is worth a post beyond "here's how to add a connector"

If you've built or used more than one MCP server, you already know the theory: implement the spec once, any compliant host can use it. What's actually interesting here is watching that promise hold up (or not) across three hosts with meaningfully different maturity levels.

Claude's MCP client implementation has been stable for a while. Tool selection is reliable, you don't need to over-specify a prompt, and the connector setup is entirely GUI-driven.

ChatGPT's MCP support (branded "Developer Mode," previously part of what OpenAI called "connectors" until a December 2025 rename to "apps") is newer and still in beta. Tool selection is noticeably less reliable, meaning you have to write more explicit prompts to guarantee it reaches for the right tool instead of falling back to something built-in like web search.

Codex has had MCP client support in the CLI and IDE extension for a while, configured through a shared config.toml, and it's the only one of the three with zero web UI involved in setup.

Same protocol. Three different levels of client maturity. If you're building your own MCP server and wondering how much you need to defensively prompt-engineer around client-side tool selection quirks, this is a useful real-world data point.

The server side, briefly

Walter's MCP server runs as a single remote, authenticated endpoint:

https://mcp-server.walterwrites.ai/mcp

It's a streamable HTTP server (not stdio), which is the relevant detail for Codex config later. It exposes two main tools relevant here, humanize and detect, both callable the same way regardless of which client is invoking them. Same server-side logic. The differences below are entirely client-side.

Setup 1: Claude (the baseline, simplest of the three)

If you've set up any MCP connector in Claude before, this is nothing new.

Settings → Connectors → click +
Select Add custom connector
Fill in:

Name: Walter Writes AI
Remote MCP Server URL: https://mcp-server.walterwrites.ai/mcp

Click Add, sign in with your Walter account, click Allow Access

That's it. No further prompt engineering required, in practice Claude reliably picks the right tool from a normal, unscaffolded prompt like:

Humanize this draft with Walter, then run detection on the result.

Setup 2: ChatGPT (beta, more moving parts)

This is currently gated to Pro and Plus accounts on ChatGPT web (plus Business/Enterprise/Education workspaces where an admin has enabled it). A few things to know before you start:

As of mid-December 2025, OpenAI renamed "connectors" to "apps." If you're looking at older docs or screenshots and can't find "Connectors" in settings, that's why.
Developer Mode gives ChatGPT full MCP client support, meaning both read and write tool calls, not the older fetch-only connector model.

Setup:

Settings → Apps → Advanced settings
Enable Developer mode
Click Create app
Fill in:

Name: Walter Writes AI
MCP Server URL: https://mcp-server.walterwrites.ai/mcp
Authentication: OAuth

Leave the Client ID/Secret fields blank, Walter's server handles the OAuth exchange automatically once you authenticate.

Check I trust this application → Create
Sign in to Walter when prompted
Confirm the app shows as connected

One more step before it's usable mid-conversation: in chat, click + → More → enable the Walter Writes connector.

The prompting gotcha

This is the part that actually matters if you want this to work reliably. Because Developer Mode is beta, ChatGPT does not consistently infer that it should call a custom tool over falling back to something built-in. A vague prompt like "make this sound more human" might silently just get ChatGPT's own rewrite instead of an actual Walter tool call.

Fix: be explicit. Name the connector, name the exact action, specify what to return, and tell it not to use web search.

Using the Walter Writes MCP connector, run the humanize action on the
draft below, then run the detect action on the result. Return the
readability score, AI score, and a short change summary. Do not use
web search.

[paste draft]

Verbose, yes. Necessary right now, also yes. Check the tool call preview in the response to confirm Walter's tools are actually what got invoked, not a built-in fallback.

Setup 3: Codex (fastest, zero UI)

If you're already comfortable in a terminal, this is the quickest of the three.

Option A, one command:

codex mcp add walterwrites --url https://mcp-server.walterwrites.ai/mcp

Option B, edit config directly. Codex stores MCP server definitions in ~/.codex/config.toml (global) or a project-scoped .codex/config.toml (only loads for directories Codex has marked trusted):

[mcp_servers.walterwrites]
url = "https://mcp-server.walterwrites.ai/mcp"

Authenticate (same OAuth flow as the other two, just CLI-triggered):

codex mcp login walterwrites

Verify:

codex mcp list

You should see walterwrites listed as active. Inside an interactive session, /mcp shows currently connected servers too.

Codex will reach for Walter's tools automatically once configured. If you want that to happen without a confirmation prompt every time, set an approval_mode on the tool (auto, prompt, or approve):

[mcp_servers.walterwrites]
url = "https://mcp-server.walterwrites.ai/mcp"
approval_mode = "auto"

For consistent behavior across a whole project without repeating instructions every session, drop a line into AGENTS.md:

Use Walter's humanize and detect tools automatically for any content
that needs AI-detection review or humanization before it's finalized.

This matters specifically for developer-facing content, README files, doc comments, CHANGELOG entries, anything written inline in a repo that might read as obviously AI-generated and would benefit from a pass before it ships.

stdio vs streamable HTTP, since it comes up

Quick clarification since this trips people up in Codex specifically: stdio servers run as a local process, spawned and piped to directly by the client. Streamable HTTP servers (what Walter uses) are remote, accessed by URL, and support OAuth or bearer token auth. That's why Walter gets added with --url in Codex rather than a local command, it's not a process Codex spawns, it's a remote endpoint Codex calls over HTTP.

Where each one actually makes sense

Not trying to force a false equivalence here, each of the three genuinely fits a different workflow:

Claude for long-form content pipelines. Tool selection just works, so it's the lowest-friction option if you're doing this regularly and don't want to think about prompt scaffolding.

ChatGPT if your team is already standardized there and switching tools just for humanization isn't worth the context switch. Budget for more explicit prompts and occasional manual tool-call confirmation until Developer Mode matures.

Codex specifically for anything developer-facing that lives inside a repo, docs, comments, generated changelogs. It's also just the fastest to set up if you're not touching a browser at all.

Common failure modes

ChatGPT: connector doesn't show up after creating it. Start a new chat. Developer Mode changes sometimes don't reflect in an existing conversation's tool list.
ChatGPT: keeps using web search instead of Walter. Prompt wasn't explicit enough, see the gotcha section above.
ChatGPT: Developer Mode isn't in Settings at all. Confirm you're on web (not mobile) and on a Pro/Plus account or a workspace where an admin enabled it.
Codex: tool call times out. Default timeout is 60s. Bump tool_timeout_sec in the server's config block if a specific call needs more.
Codex: project-level config not loading. Only loads for directories Codex has explicitly marked trusted, check that first.
Any client: OAuth won't complete. Usually a pop-up blocker or browser extension eating the redirect. Check that before assuming the server's broken.

Closing thought

The interesting part of this isn't really "Walter now works in three places." It's that it required zero server-side changes to get there, that's the actual point of MCP as a spec. What differs entirely is client-side maturity: Claude's implementation just works, ChatGPT's needs more explicit handholding right now because it's newer, and Codex trades a GUI for two terminal commands. If you're building your own MCP server, that maturity gap across hosts is worth planning around, because your users are going to hit it regardless of how well you've built your server.

Docs for all three setups: docs.walterwrites.ai. Server: walterwrites.ai.

FAQ

Does the same account work across all three?
Yes, all three authenticate against the same Walter account via OAuth. Word balance, settings, and history carry over.

Is Codex MCP support free?
The MCP client functionality itself has no separate cost, you need an active Codex/ChatGPT account with the right access tier, plus a Walter account for the humanization credits the actual tool calls consume.

Do I need a paid ChatGPT plan?
Yes, currently. Developer Mode requires Pro or Plus individually, or a Business/Enterprise/Education workspace where an admin has turned it on.

Why does ChatGPT need more explicit prompts than Claude?
Claude's MCP implementation has been stable longer and reliably infers the right tool from context. ChatGPT's Developer Mode is newer and still beta, tool selection isn't as consistent yet, so explicit naming of the connector and action is currently required for reliable behavior.

5 Real-World Use Cases for Walter MCP in Claude

Mohab Abdelkarim — Fri, 26 Jun 2026 15:05:59 +0000

Most MCP tool breakdowns stop at "here is what it does." This one goes further.

Walter MCP is an MCP server that connects Walter Writes AI directly inside Claude, bringing AI humanization, detection, and batch processing into your conversations without switching tools. If you have seen it mentioned but never understood why someone would actually build a workflow around it, this article is for you.

Below are five use cases that are working in production right now, with the logic behind each one and the prompts that make them run.

What Walter MCP Actually Adds to Claude
Quick Setup
Use Case 1: SEO Content Pipeline with Keyword Lock-In
Use Case 2: Agency QC Before Client Delivery
Use Case 3: Programmatic SEO at Scale
Use Case 4: Content Repurposing Across Formats
Use Case 5: Brand Voice Consistency Across a Team
Adding Walter Skills for Repeatable Workflows
Installing the Skills Library
Conclusion
FAQs

What Walter MCP Actually Adds to Claude

Before the use cases, a quick orientation on what the connector provides.

When you connect Walter MCP to Claude, three tools become available in every conversation:

humanize rewrites AI-patterned text to remove the statistical signatures that detection tools flag: uniform sentence rhythm, predictable transitions, vocabulary predictability. Keywords, links, brand names, and heading structure stay intact when you specify them.

detect scores text from 0 to 100 and returns a verdict (likely_human, mixed, likely_ai) with paragraph-level feedback. It tells you which specific paragraphs triggered the score and the pattern that caused it, not just a single document number.

batch_humanize processes up to 25 items in one call, each with optional individual settings. Built for content operations running at volume.

These tools run inside Claude conversations through the Model Context Protocol. Claude makes the call, the tool returns the result, and the workflow continues in the same thread. No copy-paste. No tabs.

Quick Setup

Three minutes, no developer configuration.

1. Open Claude, go to Settings > Connectors.

2. Click + then Add custom connector.

3. Fill in:

Name: Walter Writes AI
Remote MCP Server URL: https://mcp-server.walterwrites.ai/mcp

4. Click Add, then connect your Walter account and click Allow Access.

Done. Walter is now available in every Claude conversation.

Use Case 1: SEO Content Pipeline with Keyword Lock-In

The problem: You are drafting blog posts with Claude and the output is solid structurally, but it scores 90+ on AI detection and the humanizer you use keeps paraphrasing your target keyword.

Why this matters: For SEO content, the exact phrasing of your target keyword is not interchangeable. "Best CRM for small business" and "top CRM for small businesses" are different strings with different search behavior. Any humanizer that treats your content as a blob will rewrite those phrases without awareness of what it is costing you.

The Walter MCP approach:

Run detection first to identify which sections are actually flagging, then humanize only those with explicit constraints.

Step 1 — Diagnostic:
"Use Walter to detect which paragraphs in this draft are flagging as
AI-generated. Do not rewrite anything yet. Show me the paragraph-level
feedback."

Step 2 — Targeted humanization:
"Humanize only the flagged sections. Apply these constraints:
- Keep 'best CRM for small business' exactly as written in every occurrence
- Do not change any anchor text or URLs
- Preserve all statistics
- Keep the heading structure intact"

Step 3 — Verify:
"Run Walter detection on the humanized version and show me the
before/after scores."

What you get: A detection compliance report showing the score drop, a preservation compliance report confirming every keyword survived, and a humanized draft ready to review. All in one conversation.

Typical result: Raw Claude drafts that score 90 to 98 on detection consistently land in the 20 to 30 range after one targeted pass. Keyword survival rate is 100 percent when constraints are specified.

Use Case 2: Agency QC Before Client Delivery

The problem: You run an agency. Content goes through writers, gets AI-assisted, gets edited, and then goes to a QC step before client delivery. The QC step is currently manual and inconsistent: different reviewers catch different things, some keyword checks get missed, and the turnaround before delivery is longer than it should be.

Why this matters: An inconsistent QC process is a client relationship risk. If one client's content goes out with a paraphrased brand name or a detection score that surprises them, that is a credibility problem.

The Walter MCP approach:

Load this as your QC prompt at the start of each review session:

"I am going to paste a client article below. Run the following QC pipeline:

1. Use Walter to detect the full article and return the score, verdict,
   and paragraph-level breakdown.
2. For any paragraph scoring above 40, humanize just that paragraph.
3. After humanizing, run detection again on the full article.
4. Return: original score, final score, list of changed paragraphs,
   and a preservation report confirming the following survived:
   [client brand name], [target keyword], [specific product names],
   [all links].

Here is the article: [paste]"

What you get: A structured QC report for every article. Original detection score, final score after targeted humanization, list of specific changes made, and confirmation that every protected element is still intact. That report is also your documentation if a client ever asks what QC looks like on their content.

Reality check: This workflow is faster than manual QC but it is not a replacement for a final human read. Detection and preservation are mechanical checks. Whether the article makes a good argument and is factually accurate is still on you.

Use Case 3: Programmatic SEO at Scale

The problem: You are generating location pages, product description pages, or any other template-driven content at volume. You have 50 city pages to generate, humanize, and verify before they go live.

Why this matters: Programmatic SEO content produced at volume tends to have two failure modes. Either it is too uniform (detection tools treat similar pages as a pattern) or the humanization step introduces enough variation that the target keyword changes across pages. Both are problems.

The Walter MCP approach:

This is where batch_humanize becomes the key feature. One call handles up to 25 items simultaneously.

"I have a list of city-and-service combinations for local landing pages.
For each one:

1. Write a 300-word local landing page targeting '[service] in [city]'.
2. Humanize each page with Walter.
3. Preserve the exact phrase '[service] in [city]' in the H1 and
   first paragraph of every page.
4. Return a per-page detection score and preservation confirmation.
5. Flag any page that scores above 30 for my review.

Here are the first 25 city-service combinations: [list]"

Run this in batches of 25. Review flagged pages individually. Export the compliant ones.

For developers building content pipelines: The Walter MCP batch tool through Claude is a useful prototyping layer before you commit to direct API integration. You can validate your keyword constraints and output format on real content without writing any code.

The SK-03 Local SEO Machine skill from the Walter Skills library automates this exact pattern as a project-level instruction, so every chat in a project runs it without re-specifying the workflow each time.

Use Case 4: Content Repurposing Across Formats

The problem: You have a 2,000-word pillar article that took real research and effort to produce. You want to get a LinkedIn post, an email newsletter intro, a Twitter/X thread hook, and a short blog summary out of it without rewriting each format manually from scratch.

Why this matters: Content repurposing is one of those tasks that sounds simple and is actually tedious. Each format has different length expectations, different tone conventions, and different structural requirements. And if the source article was AI-assisted, every format you derive from it inherits those AI patterns.

The Walter MCP approach:

"Take the article below and produce the following four formats:

1. LinkedIn post (200 to 250 words, professional but conversational)
2. Email newsletter intro (150 words, direct, first-person)
3. Twitter/X thread hook (first post only, under 280 characters,
   strong opening)
4. Blog summary (100 words, third-person, for a related article
   roundup)

After generating all four:
- Humanize each one with Walter independently
- Run detection on each and return the score
- Preserve these data points in every format that uses them:
  [specific stat 1], [specific stat 2]
- Preserve these brand mentions: [brand name]

Here is the source article: [paste]"

What you get: Four formats, each humanized to the appropriate tone for its platform, with detection scores confirming each one reads as human-written. The data points and brand mentions you care about carry across all four.

Note on format: LinkedIn and email copy humanizes differently than blog content. The Walter humanization engine adjusts for length and register. Short-form content in particular benefits from the balanced mode rather than enhanced, which can over-engineer very short pieces.

Use Case 5: Brand Voice Consistency Across a Team

The problem: You have three writers using Claude to draft content for the same brand. Each writer prompts differently. Each gets slightly different output. The content is technically fine but does not sound like it came from the same company. The brand guidelines document exists but nobody reads it before every prompt.

Why this matters: Brand voice inconsistency compounds. The longer a team produces inconsistent content, the harder it is to establish the coherent brand identity that drives recognition and trust.

The Walter MCP approach:

This use case is best handled through a Walter Skill rather than a per-conversation prompt, specifically SK-08 Brand Voice Adapter. But you can also build a version of it inline:

"You are writing for [Brand Name]. Apply the following brand voice rules
to every piece of content in this conversation:

Brand voice: [direct / warm / technical / casual / formal]
Always use: 'platform' not 'tool', 'team' not 'staff'
Always capitalize: [Brand Name], [Product Name]
Avoid: passive voice, hedge words like 'perhaps' and 'possibly',
       sentences longer than 25 words in CTAs

After drafting any piece of content:
1. Run Walter detection and return the score
2. Humanize if the score is above 35
3. Apply the brand voice constraints above to the humanized version
4. Confirm all capitalization rules are intact before returning output"

Every writer on the team pastes this at the start of their session. Every Claude conversation in the project applies the same rules. Detection and humanization are baked in.

For consistent team use: Load the SK-08 Brand Voice Adapter skill into a Claude project's instructions once. Every writer who opens a chat in that project gets the brand voice rules and the Walter workflow applied automatically, without pasting anything.

Adding Walter Skills for Repeatable Workflows

All five use cases above work as one-off prompts. But if you are running any of them regularly, you do not want to re-specify the workflow every time you start a new conversation.

Walter Skills solve this.

A skill is a markdown file you paste once into a Claude project's instructions. Every chat in that project then applies the full workflow automatically. No re-prompting. No forgotten constraints. No inconsistent output because someone used a slightly different version of the prompt.

The full library has 12 skills:

SK-01 SEO Content Writer
SK-02 Agency QC Pipeline
SK-03 Local SEO Machine
SK-04 Content Repurposer
SK-05 E-commerce Engine
SK-06 Newsletter Writer
SK-07 Programmatic SEO
SK-08 Brand Voice Adapter
SK-09 Documentation Writer
SK-10 Lead Magnet Creator
SK-11 Social Media Suite
SK-12 Content Refresh Engine

The repo is at: github.com/walterwritesai/walter-skills

Installing the Skills Library

The Walter Skills library is on the skills.sh registry, Vercel Labs' npx CLI ecosystem for agent skills.

npx skills add walterwritesai/walter-skills

That installs the full library. Each install also counts toward the leaderboard ranking on skills.sh, which drives discovery in the registry. If you are building with Claude Code or another agent environment, this is the fastest way to get the skills locally.

For a community-maintained guide covering all 65 prompts across 12 workflow categories, waltermcp.com is the most comprehensive reference available (community-run, not an official Walter resource).

Conclusion

MCP connectors are only useful if there are real workflows to run through them. These five cover the use cases that show up most often in production content operations: SEO pipelines with strict keyword requirements, agency QC before client delivery, programmatic SEO at volume, multi-format content repurposing, and brand voice consistency across teams.

The common thread across all of them is that the workflow runs inside one Claude conversation. Detection, humanization, keyword protection, and compliance reporting all happen in the same thread where the draft was written.

If you are already using Claude for content work and spending time on the manual cleanup afterward, the connector is worth the three-minute setup. Start at walterwrites.ai.

FAQs

What is Walter MCP?
Walter MCP is the MCP server for Walter Writes AI. It exposes humanization, AI detection, and batch processing as callable tools inside Claude through the Model Context Protocol. Workflows that previously required multiple separate tools run inside a single Claude conversation.

Do I need coding experience to set this up?
No. Setup is three steps in Claude's Settings panel: add the connector URL, sign in to Walter, click Allow Access. No API keys, no configuration files, no terminal.

What is the difference between detect and humanize?
detect is the diagnostic. It scans text and returns a score with paragraph-level feedback on what is flagging and why. humanize is the fix. It rewrites the flagged patterns. Running detect first, then humanizing only the flagged sections, produces better results than running humanize blindly on the full document.

Can Walter MCP preserve exact keyword phrases?
Yes, when you specify them as constraints before the humanization runs. You name the exact phrases, brand names, links, and data that must survive. Walter returns a preservation compliance report confirming every protected element made it through the rewrite.

What is batch_humanize and when should I use it?
batch_humanize processes up to 25 items in a single call, each with optional individual settings. Use it for programmatic SEO pages, product description catalogs, or any workflow where you are processing multiple pieces of content simultaneously.

What are Walter Skills and how do they relate to MCP?
Walter Skills are markdown files you paste into a Claude project's instructions once. They encode the workflow: which detection threshold to use, which humanization mode, how to handle keyword constraints, what the output format should be. The MCP connector provides the tools. The skill provides the workflow logic so you do not re-specify it every conversation. Install them with npx skills add walterwritesai/walter-skills.

What detection scores should I expect?
Raw Claude drafts typically score 90 to 98. After one targeted Walter humanization pass using the detect-first approach, scores consistently land in the 10 to 20 range, classified as likely_human. Word limits per call depend on your Walter plan.

Where can I find more prompts for these workflows?
The community guide at waltermcp.com covers 65 prompts across 12 workflow categories. Community-maintained, not an official Walter resource, but comprehensive and regularly updated.

Walter Writes AI Humanizer MCP in Claude: How to Set It Up

Mohab Abdelkarim — Thu, 11 Jun 2026 23:40:03 +0000

If you use Claude to draft content, you probably know the annoying part.

Claude can write fast.

But when the draft is done, you still end up doing the same manual loop:

Copy the text.
Paste it into a humanizer.
Check it with an AI detector.
Go back to Claude.
Edit again.
Repeat.

That is not really a workflow. It is tab-switching.

This is where Walter MCP is interesting.

Walter Writes connects to Claude through MCP, so you can humanize AI text, run AI detection, and batch-process content without leaving the same Claude conversation.

In this guide, I’ll walk through how to set it up, how to test it, how to use it with Walter Skills, and how I would structure a real content workflow around it.

TL;DR

Walter MCP lets Claude call Walter Writes directly as a tool.

Once connected, Claude can use Walter to:

humanize AI-generated drafts
run AI detection
process multiple pieces of content
preserve important keywords and entities
reduce copy-paste between tools

For developers, agencies, and SEO teams, the main benefit is not just “better rewriting.”

The real benefit is workflow control.

Instead of treating humanization and detection as separate browser steps, you can make them part of the Claude workflow itself.

What is MCP?

MCP stands for Model Context Protocol.

The simple explanation:

MCP is a standard way for AI assistants to connect to external tools.

Instead of every tool needing a custom integration with every AI app, MCP gives tools a more standard way to expose functions that an assistant can call.

So in this case:

Claude is the AI assistant.
Walter Writes is the external tool.
MCP is the bridge between them.

Once Walter is connected, Claude can call Walter’s tools when your prompt needs them.

That is the difference between asking Claude:

“Make this sound more human.”

and asking Claude to use a dedicated humanization/detection workflow through Walter.

The second approach is more repeatable.

What Walter Adds Inside Claude

Walter’s MCP connector adds three useful actions inside Claude:

Humanize
This rewrites AI-patterned text into more natural writing while trying to preserve the meaning, structure, keywords, links, and important entities.

Detect
This checks text for AI-writing patterns and gives feedback on where the text still looks too AI-generated.

Batch Humanize
This is for processing multiple pieces at once, which is useful for agencies, content teams, and programmatic SEO workflows.

The important part is that these tools live inside the same Claude workflow.

You can draft in Claude, humanize with Walter, detect the result, and revise again without copying the text between multiple apps.

Before You Start

You will need:

A Claude account
Access to Claude connectors
A Walter Writes account
The Walter MCP server URL
A draft or sample text to test with

The Walter MCP server URL is:

https://mcp-server.walterwrites.ai/mcp

I would also keep these open while setting things up:

Official Walter Claude/MCP guide:
https://walterwrites.ai/humanize-ai-text-inside-claude/

Walter Skills GitHub repo:
https://github.com/walterwritesai/walter-skills

Community guide, not official:
https://waltermcp.com/

Important note: WalterMCP.com should be treated as a community guide, not the official Walter documentation.

Step 1: Open Claude Connectors

In Claude, go to:

Settings → Connectors

This is where you manage external tools that Claude can connect to.

Look for the option to add a custom connector.

Depending on your Claude interface, this may appear as a plus button or an “Add custom connector” option.

Step 2: Add Walter as a Custom Connector

Create a new connector with a clear name.

For example:

Walter Writes AI

Then paste the MCP server URL:

https://mcp-server.walterwrites.ai/mcp

Save the connector.

Claude should now know that Walter exists as an external MCP tool.

If Claude asks for permissions or approval when using the connector, approve the relevant actions when you actually want Walter to run.

That approval step is a good thing. MCP tools can perform actions, so you should stay aware of what Claude is calling.

Step 3: Test the Connection

Do not start with a full article.

Start with a small test.

Paste a short AI-written paragraph into Claude and ask:

Use Walter to detect whether this text sounds AI-generated.

Text:
[Paste your paragraph here]

If the connector is working, Claude should call Walter’s detection tool and return a result.

Then test humanization:

Use Walter to humanize this paragraph while preserving the meaning and keeping the keyword "AI image detector" unchanged.

Text:
[Paste your paragraph here]

This test matters because it checks three things:

Claude can access Walter.
Walter can process your text.
Claude understands your constraints.

Do not skip this step.

A lot of MCP setups technically connect but fail when you ask the tool to do real work.

Step 4: Try a Full Draft Workflow

Once the basic test works, try a real workflow.

Here is a prompt I would use:

I’m writing an article for Medium.

Please do this workflow:

1. Review the draft for AI-sounding sections.
2. Use Walter to humanize the sections that feel too robotic.
3. Preserve all keywords, brand names, links, headings, and factual claims.
4. Run detection again after humanizing.
5. Give me a short summary of what changed.

Draft:
[Paste draft here]

This is where Walter MCP becomes useful.

You are not just asking Claude to rewrite everything.

You are asking Claude to run a process.

That is a big difference.

Step 5: Use Walter Skills for Repeatable Workflows

The Walter Skills repo is useful if you do not want to rewrite the same workflow prompt every time.

A Claude Skill is basically a reusable instruction file.

Instead of telling Claude the same rules in every chat, you add the skill once to a Claude project. After that, the project knows how to behave for that type of work.

The Walter Skills repo includes skills for things like:

SEO content writing
agency QC
local SEO
content repurposing
e-commerce product writing
newsletters
programmatic SEO
brand voice adaptation
content refreshes
lead magnets
social media
docs writing

The basic install flow is simple:

Open the Walter Skills GitHub repo.
Pick the skill folder that matches your workflow.
Open the SKILL.md file.
Copy the full markdown.
Paste it into your Claude project instructions.
Make sure Walter MCP is connected so Claude can actually call Walter’s tools.

For example, if you are producing SEO articles, start with the SEO content skill.

If you are reviewing content for clients, the agency QC skill may be a better starting point.

If you are updating old articles, use the content refresh skill.

The skill gives Claude the repeatable workflow.
The MCP connector gives Claude the tool access.

You need both if you want the workflow to feel automatic.

A Simple Developer Mental Model

I think about it like this:

Claude = reasoning layer
Walter MCP = tool layer
Walter Skills = workflow layer

Claude decides what needs to happen.

Walter performs the humanization and detection actions.

Walter Skills tell Claude how to behave in a specific workflow.

That combination is more powerful than a one-off prompt.

Without a skill, you have to explain the workflow every time.

Without MCP, Claude can only rewrite with its own general writing ability.

With both, you can create a more repeatable process:

Draft → detect → humanize → preserve keywords → detect again → final review

That is the workflow most SEO teams and content teams actually want.

Example: SEO Article Workflow

Here is a practical prompt for SEO content:

You are helping me prepare an SEO article for publishing.

Use Walter where needed.

Workflow:
1. Check the draft for AI-sounding paragraphs.
2. Humanize only the sections that need it.
3. Keep the target keyword exactly as written.
4. Preserve all internal links, brand names, headings, and statistics.
5. Do not change the search intent.
6. After humanizing, run detection again.
7. Give me the final version plus a short change summary.

Target keyword:
[insert keyword]

Draft:
[paste draft]

This is better than saying:

Make this sound human.

That prompt is too vague.

The better prompt gives Claude constraints and tells it when to use Walter.

Example: Batch Humanization Workflow

For agencies, batch processing is where this becomes more interesting.

Imagine you have 10 short product descriptions or 20 local SEO pages.

Instead of processing each one manually, you can ask Claude to use Walter’s batch flow.

Example:

I have a batch of short AI-generated descriptions.

Use Walter Batch Humanize to process them.

Rules:
- Keep product names unchanged.
- Keep location names unchanged.
- Keep all prices and specifications unchanged.
- Do not add new claims.
- Return the results in the same order.
- After processing, flag any description that still sounds too AI-generated.

Items:
1. [text]
2. [text]
3. [text]

For programmatic SEO or e-commerce, this can save a lot of repetitive work.

The main thing is to be strict about preservation rules.

Any workflow that touches SEO content should protect:

target keywords
product names
brand names
locations
URLs
prices
statistics
factual claims

Humanization is useful only if it does not break the information.

Example: Detection-Only Workflow

Sometimes you do not need to humanize immediately.

You just need to check what needs work.

For that, use a detection-first prompt:

Use Walter to detect AI-writing patterns in this draft.

Do not rewrite yet.

Return:
1. Overall detection result
2. The sections most likely to be flagged
3. What writing patterns caused the issue
4. A recommended edit plan

Draft:
[paste draft]

I like this approach because it avoids over-editing.

Not every paragraph needs to be rewritten. Sometimes only the intro, conclusion, or transition-heavy sections are the problem.

Detection first helps you edit more carefully.

Common Setup Issues

Here are the problems I would check first if Walter does not work in Claude.

The connector URL is wrong

Double-check the URL:

https://mcp-server.walterwrites.ai/mcp

A small typo is enough to break the connection.

Claude has not been restarted or refreshed

After adding a connector, refresh Claude or start a new conversation.

Sometimes the connector is added, but the current chat does not pick it up cleanly.

You are asking Claude too vaguely

This prompt may not trigger the tool:

Make this better.

This is clearer:

Use Walter to humanize this text and then run AI detection on the result.

Be direct.

You forgot the preservation rules

If you care about exact keywords or entities, say that.

Example:

Preserve the exact keyword "AI image detector" and do not change brand names, URLs, or statistics.

The text is too long for the workflow

If you are testing for the first time, start small.

Use one paragraph first. Then one section. Then a full article.

Debugging a 3,000-word article is annoying if you do not know whether the connector works yet.

Security Notes for MCP

MCP is powerful because it lets AI assistants call external tools.

That also means you should treat it carefully.

A few basic rules:

only connect MCP servers you trust
review what a connector can do before using it
do not paste sensitive client data unless you are allowed to
keep API keys and private credentials out of prompts
approve tool calls intentionally
use separate Claude projects for separate client workflows

For Walter specifically, the workflow is mostly about text humanization and detection, but the general MCP rule still applies:

Do not connect random tools blindly.

MCP gives AI assistants more reach. That is useful, but it also means you need better habits.

Where Walter MCP Fits Best

Walter MCP is most useful when the content workflow already starts in Claude.

For example:

writing SEO articles
rewriting AI-generated drafts
checking content before publishing
reviewing agency content
humanizing product descriptions
preparing newsletters
refreshing old blog posts
batch-processing programmatic SEO content
checking whether a draft still sounds too AI-generated

It is less useful if you only need to humanize one short paragraph once.

For that, a normal web tool may be enough.

But if you use Claude every day for content work, the MCP setup makes more sense because it removes a lot of copy-paste friction.

Why Walter Writes Stood Out to Me

Based on my testing and comparison, Walter Writes is one of the best Claude MCP workflows right now for AI humanization and AI detection.

The main reason is that it solves a real workflow problem.

Most tools make you leave the writing environment.

Walter brings humanization and detection into Claude.

That means the draft, edit, detection, and revision loop can happen in one place.

For developers and technical content teams, that matters because the best AI workflow is not always the tool with the most buttons.

It is the workflow with the fewest unnecessary steps.

Final Workflow

Here is the setup I would use:

Claude Project
  → Walter Skill in project instructions
  → Walter MCP connector enabled
  → Draft content in Claude
  → Detect with Walter
  → Humanize with Walter
  → Detect again
  → Final manual review

That gives you a repeatable system instead of a messy copy-paste process.

MCP is not magic.

But for this use case, it is genuinely useful.

If you already write inside Claude, Walter MCP is worth setting up.

Useful Links

Official Walter Claude/MCP guide:
https://walterwrites.ai/humanize-ai-text-inside-claude/

Walter Skills GitHub repo:
https://github.com/walterwritesai/walter-skills

Community guide, not official:
https://waltermcp.com/

Walter Writes:
https://walterwrites.ai/

DEV Community: Mohab Abdelkarim

One MCP Server, Three Assistants: Running Walter Across Claude, ChatGPT, and Codex

Why this is worth a post beyond "here's how to add a connector"

The server side, briefly

Setup 1: Claude (the baseline, simplest of the three)

Setup 2: ChatGPT (beta, more moving parts)

The prompting gotcha

Setup 3: Codex (fastest, zero UI)

stdio vs streamable HTTP, since it comes up

Where each one actually makes sense

Common failure modes

Closing thought

FAQ

5 Real-World Use Cases for Walter MCP in Claude

Table of Contents

What Walter MCP Actually Adds to Claude

Quick Setup

Use Case 1: SEO Content Pipeline with Keyword Lock-In

Use Case 2: Agency QC Before Client Delivery

Use Case 3: Programmatic SEO at Scale

Use Case 4: Content Repurposing Across Formats

Use Case 5: Brand Voice Consistency Across a Team

Adding Walter Skills for Repeatable Workflows

Installing the Skills Library

Conclusion

FAQs

Walter Writes AI Humanizer MCP in Claude: How to Set It Up

TL;DR

What is MCP?

What Walter Adds Inside Claude

Before You Start

Step 1: Open Claude Connectors

Step 2: Add Walter as a Custom Connector

Step 3: Test the Connection

Step 4: Try a Full Draft Workflow

Step 5: Use Walter Skills for Repeatable Workflows

A Simple Developer Mental Model

Example: SEO Article Workflow

Example: Batch Humanization Workflow

Example: Detection-Only Workflow

Common Setup Issues

The connector URL is wrong

Claude has not been restarted or refreshed

You are asking Claude too vaguely

You forgot the preservation rules

The text is too long for the workflow

Security Notes for MCP

Where Walter MCP Fits Best

Why Walter Writes Stood Out to Me

Final Workflow

Useful Links