New: every plan rewrites your agent's text
AI text that sounds like you wrote it.
reright is an AI humanizer. Your agent sends it the text it wrote for a person and gets it back without the typical signs of AI writing, sounding like you wrote it. The more you correct it, and the more of your own writing you bring in, the closer it gets to how you write.
Works with Claude Code, and with Claude.ai and ChatGPT through a connector.
- Kind
- Goes to
- [email protected]
- From
- Claude Code
Hi Marta, the invoice export you asked for is done. You'll find it in the shared folder. Let me know if anything in it looks off.
Dear Marta, I hope this email finds you well. I am writing to let you know that the invoice export you requested has been completed and is now available in the shared folder. Please do not hesitate to reach out with any questions.
Dear Hi Marta, I hope this email finds you well. I am writing to let you know that the invoice export you requested has been completed and is now available asked for is done. You'll find it in the shared folder. Please do not hesitate to reach out with any questions. Let me know if anything in it looks off.
"Check before it's used" is on, so this rewrite waits for you before Claude Code sends it.
Try it now. No account needed.
Paste up to 300 words an AI wrote and see how reright rewrites them. You get 3 tries a day.
Your text is rewritten on Modal's servers in the US. reright does not keep it, and Modal deletes its copy within 7 days.
Starting the rewriter. The first rewrite after a quiet spell takes about 20 seconds.Rewriting your text.You have used today's 3 tries. Sign up to get 3,000 words a month free.The demo has reached its limit for today. Sign up to get 3,000 words a month free.The demo takes up to 300 words. Shorten your text and try again.The check that you are a person did not pass. Try again.The demo is busy right now. Try again in a minute.Something went wrong. Try again in a minute.reright kept your text as it was, because the language is not supported yet.reright kept your text as it was, because the rewrite took too long. Try again in a minute.reright kept your text as it was, because the rewrite did not pass its checks.N tries left today1 try left todayNo tries left todayHow it works. From your agent's text to your words.
-
Your agent writes something
Before it sends an email, a message, a commit or a PR, it calls reright with the words it wrote. Links, names, versions and code stay as they are, and fixed text such as a PR template is kept out of the rewrite.
-
reright rewrites it like you
The text comes back sounding like you wrote it. reright checks that the numbers, names, links and code match, that the language is the same and that the length is close. If a check fails, the agent gets its own text back.
-
The agent uses it
The agent checks that the meaning did not change and uses the text exactly as it came back. If the meaning changed, it keeps its own text. To see each rewrite first, turn on "Check before it's used" and approve, edit or reject it yourself.
It learns how you write. The more you use it, the closer it gets.
-
Bring in your own writing
On the Your writing page, add mail you sent from Gmail, your Google Docs, your commits and pull requests on GitHub, or files. reright uses them to match how you write and leaves out anything that looks AI-written. You do not write prompts or fill in a style guide.
-
Every correction teaches it
When you edit a rewrite before it is used, reright keeps your version to learn from, and the next rewrites get closer to how you write. Turning on "Check before it's used" for a while is the fastest way to teach it, because you see every rewrite.
-
It works where your agent writes
Claude Code connects with one line from Settings. Claude.ai and ChatGPT connect with the connector address. In Claude Code, a hook blocks git commits in shared repos, gh PR and issue text and Gmail sends whose text did not come back from reright.
Get started. It takes a couple of minutes.
-
Sign up
Create an account on the free plan with Google or an email address. You do not need a credit card.
-
Open Settings
Under Connect, copy the connector address, or generate a setup code for Claude Code.
-
Connect your assistant
Add the address as a custom connector in Claude.ai or ChatGPT, or paste the setup line into Claude Code.
-
Ask it to write something
Ask for an email or a commit message. It comes back rewritten to sound like you.
-
Add your writing
This step is optional. Bring in your mail, documents or GitHub history so rewrites start closer to how you write.
The Claude Code installer changes only files in your home directory, backs each one up first, and reright uninstall puts them back.
Pricing. Start free, move up when you pass 3,000 words a month.
Every plan has the same features, so there is no trial. Plans differ in how many words you can send a month. Text is kept 30 days on every plan.
-
Free
$0
3,000 words a month
Text kept 30 days
Learns how you write
Start free -
Starter
Recommended$5 a month
50,000 words a month
Text kept 30 days
Learns how you write
Choose Starter -
Pro
$25 a month
1,000,000 words a month
Text kept 30 days
Learns how you write
Choose Pro -
Unlimited
$100 a month
Unlimited words, fair use
Text kept 30 days
Learns how you write
Choose Unlimited
- Words are counted in the text your agent sends to reright, not in the template around it or in the rewrite.
- When a text has more words than you have left, your agent gets it back unchanged with a link to change plans.
- Months are calendar months in UTC.
- Prices are in US dollars. VAT and other taxes are added at checkout where they apply.
- Upgrades take effect at once and downgrades at the end of the billing period.
How it compares. Same texts, checked the same way.
We ran reright on the AI-written texts of the AI Humanizer Benchmark and checked its output next to the outputs of the other humanizers that the benchmark tested.
| reright | Other humanizers | |
|---|---|---|
| Changed a number, name or quote | 1% of texts | 30% to 46% (9 tools) |
| Meaning kept, judge score out of 5 | 4.41 | Lower for 8 of 9 tools (2.46 to 4.20) |
| Learns from your own writing | Yes: your mail, documents, GitHub and edits | 2 of 12 tools build a style from sample text |
| Checks facts before it returns text | Yes: numbers, names and meaning, and quotes stay as written | 1 of 12 tools says it checks numbers and meaning |
| Cheapest plan billed monthly | $5 a month for 50,000 words | $9.99 to $29.99 a month (11 tools with a paid plan) |
reright keeps quotations, greetings and headings exactly as written. When it cannot rewrite a text without changing a number or a name, it gives you your text back unchanged, which happened for 4% of the texts.
The benchmark page has the full results, including the four open AI detectors, where one tool passes more often than reright.
Texts and competitor outputs come from the AI Humanizer Benchmark, September to November 2026, run with default settings. The benchmark is run by the team behind UndetectedGPT and did not review this test. reright's outputs come from production on the evening of 6 November 2026. Facts are checked with the checker from the paper "Humanized but Still True?", and meaning is scored by the open model gpt-oss-20b. Plans and features are taken from each tool's public pages on 6 November 2026.
Your text. How long we keep it, and who can read it.
Names, phone numbers, addresses and other personal details are swapped for made-up ones before the model sees your text, and swapped back after. Amounts and dates such as deadlines stay as they are, so reright can check that the rewrite kept them. Delete your writing and everything reright learned from it is deleted too.
Rewriting runs on Modal, in the US.
Text is deleted from the live database 30 days after the decision, and from every backup within 48 hours after that.
Text is kept 30 days on every plan. Rewrites that nobody decides expire on the same schedule. Counts and billing records stay.
Text is encrypted at rest and in transit. Stored text is encrypted with AES-256-GCM under a separate key for each user, and a master key kept outside the database wraps those keys.
This is not end-to-end encryption. The server decrypts your text to show it to you and to check that a text came back from reright, and interpt runs the server. The encryption protects text in a stolen database or backup, not against the operator.
Backups reach back about a day. We can restore the service after a failure, but not text that was deleted more than a day or so ago. The privacy policy has the details.
Questions. Short answers on rewriting, enforcement, data and billing.
Does reright detect AI text?
No. It does not flag or score text as AI. It rewrites what your agent wrote so it sounds like you. The one place reright looks for AI text is in the writing you bring in: samples that look AI-written are left out, so it learns from your own words.
Will it beat AI detectors?
reright makes no such promise. The point is that the text reads like you.
What if a rewrite changes what the agent meant?
reright checks that the numbers, names, links and code match the agent's text, and returns the agent's text when they do not. The agent then compares the rewrite with what it wrote. If the meaning changed, it calls keep_original and uses its own text. If you want to see every rewrite before it is used, turn on "Check before it's used" in Settings.
Can my agent get around it?
Yes. In Claude Code a hook blocks git commits in shared repos, gh PR and issue text and Gmail sends unless that exact text came back from reright in the last 24 hours. But the hook runs on your machine, so anyone who can edit its files or set RERIGHT_BYPASS can switch it off, and it lets through text sent by a script, by curl or through the browser. Claude.ai and ChatGPT have no hook, so they call reright only because the instructions ask them to. The docs list what gets past it.
Which assistants does it work with?
Claude Code, set up with one line from Settings, which adds the hook, the MCP server and the rules text. Claude.ai and ChatGPT, added as a custom connector with the address from Settings. Your agent can send commit messages, PR text, client emails, support replies, posts and reports: anything someone else could read under your name.
What counts as a word?
A word is a run of letters or digits, so "don't" and "1,000" are one word each. In Chinese and Japanese, every two characters count as one word. Each text counts once, in the month it was sent, and waiting on it or checking it again is free. A text that comes back unchanged because the rewrite failed, after a timeout for example, does not count. A text bigger than what you have left is never cut: your agent gets it back unchanged with a link to change plans. You can still check rewrites that are already waiting.
Who can read my text?
You can. The server decrypts it for you and to check that a text came back from reright. Nobody at interpt reads it as part of running the service, but an operator with full access to the server could, which is why we do not call it end-to-end encryption.
Does the site track me?
No. This site sets no cookies and loads no analytics or third-party fonts. When you click into the try box, it loads Cloudflare's Turnstile check, and app.reright.it sets a cookie that counts your tries for 2 days. Your theme choice is saved in your browser and never sent anywhere. The app at app.reright.it uses a session cookie so you stay logged in.
Who do I pay, and who sends the invoice?
interpt sells reright directly and issues the invoice. Stripe takes the card payment and calculates the tax at checkout. Card details go to Stripe and never reach us.
How do I cancel?
Downgrade to Free in Settings. It takes effect at the end of the period you paid for. Deleting your account removes your data and cancels the paid plan.