HumanizeFlow for Agents

AI humanizer MCP.
Your writing workflow, connected.

Send an existing English draft from your Agent. Get a careful rewrite, protected-detail checks and meaning review, using the same Pro allowance as the website.

MCP tools require an active Pro membership. Pro Pilot enrollment is not open yet.

Connect once.
Edit where you write.

Use a remote MCP client that supports HTTP and OAuth authorization.

  1. Add this URL as a remote MCP server in your Agent.
  2. Sign in to HumanizeFlow and review the requested scopes. An active Pro account is required to use tools.
  3. Authorize the connection, then ask your Agent to check your allowance and edit a complete draft.

OAuth uses a scoped connection. Keep access tokens in your client’s credential store. Revoke access from Your account when you stop using a client.

One draft.
A useful edit.

The Agent keeps your article-generation workflow. HumanizeFlow handles the editing step.

Use HumanizeFlow to check my remaining allowance. Light-polish this complete English blog section, preserve the links and quoted phrases, and keep the argument unchanged. Show the revised text and any review flags before I publish it.

Pass the draft, purpose, mode and exact protected phrases. Confirm the protection review before generating. Keep a stable request ID for this intended edit so a retry can recover the same task.

Returned text is accompanied by preservation checks, meaning-review status and remaining usage. Read the result in context; an automatic review does not independently verify your original facts.

Tools your Agent can call.

humanize_text

Edit a complete draft

Choose Email, Blog, Marketing, Academic or Business context. Use light polish or deep rewriting and specify phrases that must remain.

refine_sentences

Improve a connected passage

Request up to two checked alternatives for one to three adjacent sentences. Selecting a returned option uses no additional allowance.

get_usage

Check your shared allowance

Read remaining words and the per-draft limit before generating. Website and Agent calls draw from the same subscription period.

get_run

Recover an existing result

After a timeout, look up the original request. Completed results are available for 15 minutes without another rewrite charge.

cancel_run

Stop a pending request

Release a pending reservation. A request that has already completed remains charged.

View tool input schemas and example requests ↗

One Pro balance.
Two ways to use it.

Choose the website or your Agent for each editing task.

Pro includes 50,000 successful input words per subscription period, with 50–5,000 words per draft. A successful 3,000-word Agent rewrite leaves 47,000 words for either entry point when starting from a full balance.

Failed or canceled requests do not consume word allowance. Queries and result recovery are free. One pending rewrite per account prevents conflicting reservations. Sentence alternatives are charged once against their full review context, not only the selected sentence.

Check the balance before a long task. Use the same request ID after a timeout and recover the existing result instead of repeatedly starting new edits.

Start from the writing task.

Explore the same editing contexts in the website workbench.

EmailMake the request clear. Keep the recipient, deadline and level of commitment.Blog & webMake articles easier to follow. Preserve your argument, headings and supporting details.MarketingMake the value clear. Preserve the offer, product facts and conditions.AcademicImprove clarity while retaining terminology, citations and the strength of your evidence.BusinessMake reports and updates direct. Preserve responsibilities, dates and qualifications.

MCP questions,
answered.

Connection, allowance and editing boundaries.

What is an AI humanizer MCP server?

It exposes editing tools to an MCP-capable Agent. Your Agent can submit an existing English draft, choose an editing context and receive revised text with protection and meaning-review information. HumanizeFlow does not generate a new article from scratch.

Do website and MCP calls share credits?

Yes. An active Pro account has one allowance of 50,000 successful input words per subscription period, with 50–5,000 words per draft. Website and MCP calls use that same balance. Failed and canceled requests release their reservations; read-only tools are free.

How is sentence refinement charged?

A successful candidate batch charges the larger word count of the full original draft or full working text once because each alternative is reviewed in context. Choosing an option is free. No safe candidates means no word charge.

Can I connect Claude, Cursor or another Agent?

Use a client that supports remote MCP over HTTP with OAuth authorization. Configure the server URL below and follow your client’s connection flow. Client configuration and account support vary; this page does not claim every version has been tested.

What happens if my Agent retries after a timeout?

Keep the same request_id and identical inputs for retries. A completed request is not charged again. Use get_run with that ID to recover its result within 15 minutes. A new ID starts a new intended request, so avoid automatic repeated rewriting.

Can I use MCP without Pro?

The free website editor remains available without an account. Agent tools require a verified account and an active Pro membership. Pro Pilot enrollment is currently closed while payment setup and acceptance are completed.

Does this guarantee a human score from an AI detector?

No. The tools improve expression while checking protected details and meaning. Detector classifications do not establish authorship; preserve required attribution and AI disclosures and review the final text.

What can a connected Agent access?

Only the scopes you authorize: rewriting, shared usage and owned run status/results. It does not receive your website history, account administration or payment controls. You can revoke a connection from Your account.

Use authorized drafts and review the output. Preserve required attribution and AI disclosures. Privacy details · Responsible use