AI Development
Claude Code puts your account email in the User-Agent: why it happens and how to stop it before the fix
Claude Code can put your account email in the User-Agent of requests to websites. Anthropic removes it in the next release: until then, do these 3 things.

Luca Di Domenico
· 5 min read
In short
If you sign in with your Claude account, Claude Code passes your email to the model, and Claude sometimes writes it into the User-Agent of requests to websites without asking you: one user counted 53 calls across 8 sessions from August to November 2026. Anthropic confirmed the problem on Hacker News and will remove the email from the context in the next release. Until then, update to at least 2.1.317 (the 2.1.311 stable has no protection), don’t rely on cc-plugin-privacy alone, since it covers a single case, and add a rule to your CLAUDE.md.
Yes: Claude Code can put your Claude account email into the requests it sends to the websites it visits, without asking you. Lydia Hallie, who works on Claude Code, confirmed it on Hacker News on November 6, 2026. Anthropic will remove the email from the model’s context in the next release, expected early next week. Until then there are three things to do: update Claude Code to at least 2.1.317, don’t rely on the cc-plugin-privacy block alone, and add a rule to your CLAUDE.md.
What happened with the email in Claude Code?
On November 5, 2026 a user opened issue #107757 on GitHub, "Privacy violation in User-Agent header". They write that they get dozens of scam emails a day, because Claude Code gave their address to unknown websites while searching the web.
In the same issue another user asked Claude to check their past sessions. The result:
- 53 calls with their email inside the User-Agent;
- spread across 8 sessions, from August 9 to November 1, 2026;
- almost all of them from subagents writing their own
curlcommand with the email, not from the WebFetch tool.
The issue reached the Hacker News front page, and that is where Anthropic replied.
Why does Claude Code send your email to websites?
If you sign in to Claude Code with your Claude account, Claude Code passes the account email to the model. According to Lydia Hallie, Anthropic added it so Claude would stop inventing names in fields like @author tags.
The problem comes from the User-Agent. It is the business card every request shows to a website: it says who is asking for the page. Some sites, like Wikipedia, ask automated programs to leave a contact there.
When Claude writes its own command to download a page, it sometimes puts the email from its context in that field. The site receives it, and you never authorized it.
When will Anthropic’s fix ship?
Anthropic said it will remove the account email from the model’s context in the next Claude Code release, expected early next week. As of November 7, 2026 the fix is not out: version 2.1.319, released on November 6, does not mention it in the changelog, and issue #107757 is still open.
The problem had already been reported. Issue #81138, which reports the account email being injected into the model’s context without consent, has been open since July 25, 2026.
What should you do right now to protect your email?
1. Update Claude Code to at least 2.1.317
Check your version with:
claude --versionClaude Code’s stable channel is still on 2.1.311, and 2.1.311 has no protection at all. Starting with 2.1.317, Claude Code asks you before a web or shell tool call sends your account email to the network.
2. Don’t rely on cc-plugin-privacy alone
Version 2.1.317 added a built-in block called cc-plugin-privacy, on by default. You can find it in /plugins: check that it is there and enabled.
The block covers a single case, though: a terminal command that contains your email and names a User-Agent, like a curl call. The mod’s README lists what stays uncovered:
- web searches and page reads (WebSearch and WebFetch);
- MCP tools;
- scripts that Claude writes and then runs;
- the email placed in the page URL or in the request body.
The README itself calls it "a courtesy" that "covers one shape". And in the issue two users showed that on 2.1.318 the block doesn’t even stop that one case.
3. Add a rule to your CLAUDE.md
This is the rule I use, ready to paste into your CLAUDE.md file:
## Privacy
Never put my email in the User-Agent, headers, URLs or data of requests to websites, even if the site asks for it. If a site asks for a contact, stop and ask me.The rule tells Claude what to do when a site asks for a contact: stop and ask you, instead of deciding on its own. Together with the first two steps, it covers the cases the built-in block leaves open.
Why does it matter if you use Claude Code for work?
An email sent to strangers without consent is lost personal data. If you use Claude Code for work, and especially if you handle client data, the problem reaches your clients too. For the same reason it pays to check the tools you install next to Claude Code: I covered the 2 checks to run after the backdoor in the Claude Code extension for Cursor.
Anthropic’s fix is close, but until it ships your email stays in the model’s context. Updating, checking /plugins and adding the rule takes five minutes: do it today, and share this article with whoever on your team uses Claude Code.
Sources
- Bug Report: Privacy violation in User-Agent header (issue #107757, anthropics/claude-code) — Opened on 11/5/2026 by taylor-edwards on Claude Code 2.1.318: dozens of scam emails a day after the email ended up in the User-Agent during web searches. Includes morsik’s comment (53 calls across 8 sessions, 8/9 to 11/1/2026) and poteat’s comment on the cc-plugin-privacy mod. Still open on 11/7/2026.
- Hacker News: Claude Code sends your email address to websites — Comment by Lydia Hallie (works on Claude Code), 11/6/2026: the email is removed from the model’s context in the next release, expected early the following week; it had been added to stop invented names in @author tags; the stable channel is still on 2.1.311, update to 2.1.317 or later.
- README della mod privacy (anthropics/claude-code, mods/privacy) — Read on 11/7/2026: the cc-plugin-privacy mod only refuses a Bash, PowerShell or Monitor call that holds the email and names a User-Agent; it does not check WebFetch, WebSearch, MCP, scripts, variables, or the email in a URL or body; "a courtesy" that "covers one shape".
- Claude Code changelog — 2.1.317 (11/3/2026): "asks before a web or shell tool call sends your account email address to the network". As of 11/7/2026 the latest version is 2.1.319 (11/6), with no entries about the email or the User-Agent.
- Account email (PII) is injected into the model's system prompt without consent (issue #81138, anthropics/claude-code) — Opened on 7/25/2026 and still open: the OAuth account email enters the model’s context in every session, with no setting to turn it off.
Want to know what this means for your product?
If you're building software with AI and want a second opinion on how you're working, tell me about your project.
Tell me about your projectFollow the blog
One email for every new article
I only write when an article is out: title, summary and link. You can unsubscribe with one click from any email.
RSS feed
An RSS feed is the up-to-date list of articles: add it to a reader app like Feedly or Inoreader and new articles show up there, no email needed.