Claude Code "organization has disabled Claude subscription access"
What Claude Code's "organization has disabled subscription access" and "usage allocation disabled by your admin" mean, who can change them, and the server case.
“Your organization has disabled Claude subscription access for Claude Code” means the Claude organization you signed in to does not let its members use Claude Code with their subscription login. It is set on Anthropic’s side, so /login with the same account, local settings and flags change nothing; only an admin of that organization can turn it back on. “Your usage allocation has been disabled by your admin” is a different control: your sign-in works, but the spending is refused, which by the refusal’s own field reads as the admin having set your usage-credit limit to zero; an Owner changes that under Organization settings > Usage. Until then, ask the admin, or use a credential that is genuinely yours for work that is yours. A scheduled job on a token from that organization fails on every run until the admin acts.
“Organization has disabled Claude subscription access”
The full line, as Anthropic’s error reference gives it (November 2026):
Your organization has disabled Claude subscription access for Claude Code · Use an Anthropic API key instead, or ask your admin to enable accessCopiedThe reference says the organization “doesn’t allow signing in to Claude Code with a subscription login”, that running /login again with the same account returns the same error, and that it “can’t be overridden from local settings, environment variables, or CLI flags.” In claude -p and the Agent SDK the same refusal arrives as the error code oauth_org_not_allowed, which is the string to match in a script.
“Organization” here is the Claude organization the login belongs to: on a Team or Enterprise plan, the company’s. If the same email belongs to more than one organization, signing in to the wrong one is the first thing to rule out: /status shows the login the session is using, and /logout then /login lets you pick again.
Who can turn it back on
An admin of that organization. Anthropic’s fix is to “ask your admin to enable Claude Code access for your organization”, and if you are the admin and see no option for it, to contact Anthropic support. The error reference does not name the switch. The controls Anthropic does document for who gets Claude Code are these, and an admin looking into it would check them:
- The seat. Claude Code comes with every Team seat and with the single seat of new Enterprise plans; older Enterprise plans include it only on “Chat + Claude Code” and Premium seats. Owners change seats under Organization settings > Organization, per Anthropic’s Team and Enterprise article.
- The role, on Enterprise. A custom role’s Capabilities tab has a Claude Code capability, “Access to Claude Code” (custom roles).
On a personal Pro or Max account
The message also reaches people with no Team or Enterprise organization at all. Personal Pro and Max subscribers have reported it in #82700 (open since July 2026) and #68212, with no admin to ask and no explanation from Anthropic on either thread as of November 18, 2026. Nothing on the machine changes it, by the reference’s own account. The route is Anthropic support, from the account that pays for the plan; one reporter in #82700 got access back after support suggested re-syncing the Google Play subscription the plan was billed through, without being able to say which change fixed it.
“Your usage allocation has been disabled by your admin”
This one is not in Anthropic’s error reference as of November 18, 2026. The wording comes from reports by projects that run Claude Code unattended, one of them quoting Claude Code 2.1.309:
Your usage allocation has been disabled by your admin · run /usage-credits to ask your admin for a higher limitCopiedThe sign-in is accepted; it is the spending that is refused. In that report’s stream-json output the refusal carries overageDisabledReason set to member_zero_credit_limit. Anthropic’s Enterprise Admin API guide says of a member’s usage-credit limit that “‘0’ means usage credits are disabled for that member”, and the field name points the same way: the admin has set your account’s usage-credit limit to zero, so a request that would need credits has nothing to draw on. That reading is an inference from the field name; Anthropic has not documented the message.
The setting lives in Anthropic’s usage-credits article: Owners and Primary Owners open Organization settings > Usage, find Spend limits by user, and choose Edit limit from the member’s menu. Some Team organizations show Spend by org member there instead and have no per-member limits. On seat-based Enterprise, the request a member sends asks the admins either to move them to a Premium seat or to enable usage credits for their account.
The message names /usage-credits, which on Team and Enterprise sends that request to your admin for you. It is a request, not a switch: nothing changes until an admin acts on it.
Messages that sit beside these
- “You haven’t been added to your organization yet”. Anthropic’s admin setup page says the member’s seat “doesn’t include Claude Code access and needs to be updated in the admin console.”
- “You’ve hit your individual spend limit”, “your team’s shared budget”, “your individual usage limit”. A limit the admin set has been used up, rather than set to nothing. An admin raises it at the same Usage page, and where your plan’s window is what ran out, the message gives a reset time.
- “Your organization has disabled API key authentication”. The reverse: a Claude Console admin has turned off API keys, and the fix is the subscription login or the admin re-enabling keys in the Console.
- “OAuth token revoked”. The login itself was withdrawn, which the reference says happens when “an admin removed access”; the guide to Claude Code’s sign-in errors covers it.
What you can do while you wait
Send the admin the exact line, which of the two messages it is, and the machine and job it stops. The admin’s decision may be deliberate, a cost or compliance choice for the whole organization, and only they can say.
The message offers an API key, and Anthropic’s reference lists one as a fix. That is the right move when the key is one you are entitled to use for this work: a key from your organization’s own Console, issued for it, or your own Console account for your own projects. A key or a personal subscription used to carry on the organization’s work after its admin switched that work off is not a fix; it is the decision the admin made, undone without them. API usage is billed per token to the Console organization the key belongs to, not to any plan, as the plans guide explains.
If you also have a personal plan for your own projects, keep the two logins apart rather than switching back and forth. Anthropic’s authentication page gives each account its own configuration directory:
alias claude-work='CLAUDE_CONFIG_DIR=~/.claude-work claude'CopiedWhen a scheduled job on a server starts failing
A token from claude setup-token, or a /login saved on a server, belongs to the account and organization that created it, and the organization’s setting is checked on Anthropic’s side. So when an admin changes the policy, a job that worked yesterday fails today with nothing changed on the machine, and making a new token from the same account in the same organization gets the same refusal.
Neither error goes away on its own. The usage-allocation refusal in the stream-json report above carried a reset time weeks away, and the project’s fix stopped treating it as a rate limit to wait out. The same report shows the run’s final result line with "subtype": "success" and "is_error": true, so a wrapper that reads only the subtype counts the failure as a success. Look for the text instead, and tell yourself:
#!/bin/sh
# Run a claude -p job; stop and report on an admin refusal.
out=$(claude -p "$1" 2>&1)
status=$?
if printf '%s' "$out" | grep -qE \
'oauth_org_not_allowed|subscription access|allocation has been disabled'
then
printf '%s\n' "$out" | mail -s "claude: blocked by org admin" you@example.com
exit 1
fi
exit $statusCopiedReplace the address and the mail command with whatever reaches you. A job that runs every few minutes otherwise just fails every few minutes, and running Claude Code headless on a schedule covers the rest of hearing back from one.
If the admin agrees the job should move to an API key, remember the order Claude Code picks credentials in: ANTHROPIC_API_KEY outranks CLAUDE_CODE_OAUTH_TOKEN and a saved login, and in -p mode “the key is always used when present.” Remove the old token from the job’s environment when you add the key, so the next person reading it knows which credential the job runs on. A key from a Console organization with no credit fails differently, as the credit balance guide describes.
The same shape exists on OpenAI’s side, where a ChatGPT workspace admin can refuse a sign-in method for Codex: the workspace device-code guide covers that one.