How to Connect Phrasly to Claude and Humanize Text Without Leaving the Chat

If you draft in Claude, the humanizing step has always meant leaving the conversation: copy the text, open another tab, paste it, wait, copy the result back. The Phrasly connector removes that round trip. Connect it once and you can ask Claude to humanize or AI-check anything, right where you wrote it.

This guide covers what the connector does, how to add it in every Claude app, the prompts that get the best results, and how to fix the few things that can go wrong.

What the connector actually does

Phrasly runs a remote MCP server. MCP is the open standard Claude uses to talk to outside services, and connecting one gives Claude a set of tools it can call on your behalf. The Phrasly connector gives it three:

You never type a tool name. You ask in plain language and Claude picks the right one. Everything runs against your Phrasly account, using your plan and your word balance, and every humanization is saved to your Phrasly document history so you can reopen it on the website later.

Before you start

• A Phrasly account. Free accounts work, and paid plans bring their full word balance with them.

• A Claude plan that supports custom connectors: Pro or Max for individuals, or Team and Enterprise where an owner adds it for the organization. Claude Code connects on any plan that includes it.

The connector URL is the same everywhere:

https://phrasly.ai/api/mcp

Set it up on Claude.ai or Claude Desktop

1. Open Settings and go to Connectors.

2. Click Add custom connector.

3. Give it a name, paste https://phrasly.ai/api/mcp as the URL, and click Add.

4. Click Connect. A browser window opens on phrasly.ai.

5. Sign in to Phrasly if you are not already, then review what the connector is asking for and choose Allow.

6. You are returned to Claude and the connector shows as connected. That is it.

   

On Team and Enterprise, an owner adds the connector once for the organization. Each member still clicks Connect and signs in with their own Phrasly account, so word balances and history never mix between people.

Set it up in Claude Code

One command:

claude mcp add --transport http phrasly https://phrasly.ai/api/mcp

Then start Claude Code, run /mcp, select phrasly, and choose Authenticate. Your browser opens the same approval screen. Once you approve it, the tools are available in that project.

Using it: prompts that work

Ask for what you want in normal language. These all work:

• "Humanize this so it does not read as AI: ..."

• "Rewrite the intro I just wrote with Phrasly, then show me the AI score before and after."

• "Check whether this paragraph reads as AI-written."

• "How many Phrasly words do I have left?"

Phrasly also registers three slash commands. In Claude Code they appear as /mcp__phrasly__humanize, /mcp__phrasly__humanize-and-verify, and /mcp__phrasly__check-ai. Each one defaults to the most recent draft in the conversation, so you can write something and humanize it with a single command.

Get proof it passes

Ask Claude to humanize and verify, and the result comes back with a receipt: the AI score before, the score after, and the words charged. It is the fastest way to show a client, an editor, or yourself that the final text reads as human.

Long documents

There is a per-request limit of 800 words on free and trial accounts and 5,000 words on paid plans, so a long piece has to be humanized in chunks. Claude handles this well if you ask it to: "humanize this section by section, keeping the headings intact." Work through a long document a few sections at a time rather than pasting the whole thing at once.

Editing after humanizing

If you want changes to the humanized text, ask Claude for them, then ask it to run the edited version through Phrasly again. This matters more than it sounds: edits written by an AI can reintroduce the patterns detectors look for, so a final pass through the humanizer keeps the text clean.

Troubleshooting

Claude says it needs human verification

Phrasly runs occasional human checks to keep automated abuse off unlimited plans. When one is due, the tool hands back a link to phrasly.ai/human-verification. Open it in your browser, complete the short check, then tell Claude to try again. The check exists to confirm a person is present, so complete it yourself rather than asking Claude to do it. It applies to the website as well, and clearing it in one place clears it in both.

"Out of words" or an upgrade prompt

The connector enforces exactly the same plan rules as the website. If you are out of words, top up or upgrade at phrasly.ai and the tools start working again immediately. Ask Claude for your account status any time to see where you stand.

The connection fails or the tools disappear

Access is renewed automatically in the background, but a connection can expire if it has gone unused for a long time, or if you revoked it. Reconnect from Claude's connector settings and sign in again. Reconnecting also refreshes the tool list, which is worth doing if Phrasly ships new tools.

Text comes back unchanged or too short to process

Humanization needs at least 30 words, and AI detection needs at least 50, because scores on very short passages are unreliable. If Claude reports the text is too short, give it more context or humanize a larger section.

Privacy and control

• The connector only ever acts on the Phrasly account you signed in with, and only with the permissions shown on the approval screen.

• You can revoke access at any time from your Phrasly account settings. Revoking takes effect immediately.

• Humanizations are saved to your Phrasly document history so you can reopen them on the website. If you have turned off document storage in your preferences, that preference is respected here too.

Frequently asked questions

The answers to the most common questions are collected at the bottom of this page. In short: the humanizer is the same one the website uses, your words are billed the same way, your work is saved the same way, and you stay in control of the connection.