Claude Enterprise Support: Anthropic, AWS, Google 2026

TL;DR

Claude enterprise support from Anthropic is written only: messenger and email, no phone, no published response times. Bedrock, Google Cloud and Foundry customers open cases with the cloud provider, which publishes targets of 15 to 60 minutes for critical cases on paid plans. None of the three documents how a model fault reaches Anthropic.

What support does Anthropic give each Claude plan?

Anthropic offers written support through an in-app messenger and email, with an AI bot called Fin in front. Pro, Max, Console admins and Team or Enterprise owners can reach a human. Ordinary Team members cannot. Anthropic publishes no response-time target for any plan and offers no phone line.

The Claude support guide is blunt: “Support is provided in writing through the support messenger and email. We don’t offer phone support.” Replies arrive asynchronously in the same conversation and by email. On response times it says only that they “vary by plan and by the severity of the issue.” There is no number to put in a contract annex.

Route	First contact	Human access	Published response target	Extra cost
Free	Fin bot, help docs	No	None	None
Pro, Max	Fin, then Product Support	Yes, async messenger	None	None
Team	Fin for all members	Only Primary Owner, Owner, Console Admin	None	None
Enterprise	Fin	Owners plus designated support contacts, prioritized options on qualifying plans	None	Customer success only at spend thresholds (sales-assisted)
Claude API (Console)	Fin	Console admins via Console messenger	None	None
Bedrock	AWS Support	AWS engineers	AWS plan targets	AWS support plan
Google Cloud	Google Cloud Customer Care	Google engineers	Google plan targets	Google support plan
Foundry	Microsoft Support	Microsoft engineers	Azure plan targets	Azure support plan

The Enterprise difference is about who may ask, not how fast someone answers. Owners can designate support contacts, including whole groups, so a service desk can talk to Anthropic without holding an Owner role. Owners and admins on “qualifying Enterprise plans” get prioritized options in the messenger. The pricing page lists “customer success support at certain spend thresholds” only for sales-assisted Enterprise, not for self-serve Enterprise or Team. Live working sessions, per the support guide, go through the Anthropic account team.

On the API side, the one documented availability number is gone for new buyers. The service tiers page says Priority Tier, which targets 99.5% uptime, is no longer for sale. Existing commitments run to their end date. Standard tier is best effort.

What do AWS, Google Cloud and Microsoft charge for support?

All three clouds sell support as a percentage of your total monthly cloud bill, with a floor. AWS Business Support+ starts at $29 per account, Enterprise at $5,000 and Unified Operations at $50,000 per month. Google charges $29, $100 or $15,000 minimum. Azure lists $29, $100 and $1,000 flat. The fee covers Claude because Claude runs on your cloud invoice.

Plan	Minimum per month	Percentage of cloud spend	Fastest critical response
AWS Basic	Included	None	No technical cases
AWS Business Support+	$29 per account	9% up to $10K, falling to 3% over $250K	Under 30 min
AWS Enterprise Support	$5,000	10% up to $150K, falling to 3% over $1M	Under 15 min
AWS Unified Operations	$50,000	10% up to $1M, falling to 5% over $5M	Under 5 min
Google Standard	$29	3%	No P1 target, P2 in 4 h, 8/5
Google Enhanced	$100	10% up to $10K, falling to 3% over $250K	P1 in 1 h, 24/7
Google Premium	$15,000	10% up to $150K, falling to 3% over $1M	P1 in 15 min
Azure Developer	$29	Flat	Sev C in 8 business hours
Azure Standard	$100	Flat	Sev A in 1 h
Azure Professional Direct	$1,000	Flat	Sev A in 1 h

Sources: AWS plans, AWS pricing, Google Cloud Customer Care pricing, Azure support plans.

Two traps. First, AWS Basic is not a fallback for incidents. The AWS Support FAQ states that Basic customers “cannot open technical support cases”. If your Claude account on Bedrock sits on Basic, your only channel during an outage is the dashboard and re:Post. Second, the percentage applies to the whole account bill, not to Bedrock. Put Claude in a dedicated account and the Business Support+ fee on $8,000 of token spend is $720 a month (9%), our arithmetic from the published tiers.

Language matters for European teams. The AWS CreateCase API lists English, French, Japanese, Chinese, Korean, Spanish, Portuguese and Turkish as case languages. German is not on the list. Google Enhanced and Premium list English, Japanese, Mandarin Chinese, Korean and French.

Who handles a model issue on Bedrock, Google Cloud or Foundry?

The cloud provider is your only documented first contact. Anthropic sends Bedrock customers to AWS Support, Microsoft routes all Foundry questions to Microsoft Support, and Google treats Claude as a third-party “Separate Offering”. None of the three publishes how or when a case about model behaviour gets passed to Anthropic.

The Anthropic article for Bedrock users says: contact AWS Support or your AWS account manager. Only customers with a private offer and a direct Anthropic contract get pointed to an Anthropic relationship manager. Under AWS Service Terms section 50.12.1, third-party models on Bedrock are “Third-Party Content” and AWS is the invoicing party. AWS sells and bills Claude. The AWS FAQ describes its third-party software support as installation, configuration and troubleshooting for operating systems, web servers and databases, with deeper topics sent to the vendor. Foundation models are not on that list.

Google is clearer about distance. Its Claude model page states: “Anthropic models are not a Google product”. In the Service Specific Terms, Google disclaims all liability arising from use of Separate Offerings. Microsoft’s hosting comparison lists Microsoft Support as the support path for both hosting options, and adds that Anthropic “is the operator and provides any SLA”.

Issue type	Bedrock	Google Cloud	Foundry
Quota, throttling, 429 or 503 errors	AWS Support	Google Customer Care	Microsoft Support, quota form
IAM, network, endpoint, billing	AWS	Google	Microsoft
Wrong or degraded model output	AWS first, onward path undocumented	Google first, onward path undocumented	Microsoft first, Anthropic operates the model
Model availability SLA	AWS terms	Google terms, Claude excluded as Separate Offering	Anthropic, “any SLA”
Status channel	AWS Health Dashboard	Personalized Service Health	Through Microsoft Support

Our reading: infrastructure faults get a fast, contractually timed answer from the cloud. A quality regression in Claude itself lands in a queue whose second step you cannot see. Plan for that, rather than assume the cloud engineer can fix a model.

How do you escalate a Claude-on-Bedrock incident?

Open an AWS Support case with the data AWS asks for, at the highest severity your plan allows, and watch AWS Health in parallel. Without Business Support+ or above you cannot open a technical case at all. AWS’s own Bedrock guidance lists the fields that let an engineer act without a second round trip.

The Bedrock scaling guide asks for the endpoint, Region, model or inference profile ID, HTTP status and API error type, request IDs, UTC timestamps, token rate, request rate, concurrency and scaling timeline. For persistent 503 or 529 errors it says: stop ramping, check the AWS Health Dashboard, then contact AWS Support with request IDs and timestamps.

1. Confirm scope. Check AWS Health, free for all customers, and status.claude.com. The Anthropic page does not list Bedrock, so a green status there says nothing about your EU inference profile.
2. Stop the ramp. Return to the last stable request rate. Retries alone amplify load.
3. Collect evidence: profile ID such as eu.anthropic.claude-sonnet-5, Region, request IDs, UTC timestamps, error type.
4. Open the case with the right severity. Production down maps to under 1 hour on all paid AWS plans; business-critical down is faster only on Business Support+ and up.
5. Ask in the case whether AWS will involve the model provider, and request the reference. This step is not documented, so get it in writing.
6. If you hold a private offer with Anthropic, inform your Anthropic relationship manager in parallel.

Teams that open cases from their incident tooling can use the Support API, which needs Business Support+, Enterprise or Unified Operations. Look up the codes first, because AWS assigns them per service:

aws support describe-services --language en --query "services[?contains(name, 'Bedrock')]"
aws support describe-severity-levels --language en
aws support create-case --subject "Bedrock eu.anthropic.claude-sonnet-5 sustained 503 in eu-central-1" --service-code "$SERVICE_CODE" --category-code "$CATEGORY_CODE" --severity-code "$SEVERITY_CODE" --language en --communication-body "Endpoint bedrock-runtime, Region eu-central-1, profile eu.anthropic.claude-sonnet-5, HTTP 503 since 08:12 UTC, request IDs attached, 40 RPM, concurrency 12."

The setup for the profile itself is in our Claude on AWS Bedrock EU guide. On Google Cloud, the same pattern runs through Customer Care and Personalized Service Health; our Claude on Vertex AI in Europe guide covers the EU endpoint.

Run it yourself or outsource?

Claude support for a company is mostly first-level work: triage tickets from users and developers, decide whether the fault is your app, your cloud account or the model, then open the right case with the right evidence. That work stays with you even if you outsource, because only your team sees your logs and your users.

What running it in-house means, as recurring tasks:

Task	Typical owner	Frequency
Answer user questions on Claude apps, Claude Code, prompts	Service desk, trained power users	Daily
Triage: app bug vs. quota vs. model regression	Platform or cloud team	Per incident
Open and chase AWS, Google or Microsoft cases	Cloud team with case rights	Per incident
Hold Anthropic support-contact role (Enterprise)	Designated contacts	Per incident
Watch status.claude.com and the cloud health service	On-call, automated subscription	Continuous
Track model deprecations and quota changes	Platform owner	Monthly

We do not have sourced effort figures. Our estimate: below roughly 200 users, a named owner in the platform team plus a short runbook covers it; the bottleneck is case rights and evidence collection, not headcount.

Outsourcing to a managed service provider makes sense when you lack on-call coverage outside business hours, when you need German-language first-level support that the clouds do not offer, or when Claude sits in a regulated process and you need written incident timelines. It makes less sense if your cloud team already runs AWS or Google incidents 24/7: adding Claude to that runbook is cheaper than a second party in the chain.

What to demand from a provider for Claude support specifically:

• Response targets per severity, written and measured, that are no weaker than your cloud plan.
• Who opens cloud cases and on which account. Ideally your account, with the provider holding a scoped IAM role for support:* actions.
• A DPA covering ticket contents, since prompts and outputs end up in tickets, plus a current subprocessor list.
• Access model: least privilege, no standing admin rights on the Claude Enterprise org or the Bedrock account, logged break-glass access.
• Exit: runbooks, ticket history and case references handed over in a usable format on termination.

FAQ

Does Anthropic offer phone support for Claude Enterprise?

No. Anthropic states that support is provided in writing through the messenger and email and that it does not offer phone support. Enterprise customers who need a live working session arrange it through their Anthropic account team.

What does Claude support cost for a company?

Support from Anthropic carries no separate fee on any plan; customer success comes only with sales-assisted Enterprise at certain spend levels. On a cloud route you pay the cloud plan: AWS Business Support+ from $29 per account or 9% of spend up to $10K, Google Enhanced from $100, Azure Standard $100 per month.

Anthropic support vs. AWS Support: which is faster for Claude?

AWS publishes targets, Anthropic does not. AWS Business Support+ promises under 30 minutes for business-critical system down and under 1 hour for production down. Anthropic says only that response times vary by plan and severity. For contractual response times, the cloud route wins.

Who do I contact if Claude on Bedrock gives wrong answers?

AWS Support. Anthropic directs Bedrock users there unless they hold a private offer with a direct Anthropic contract. Whether and how AWS forwards a model-quality case to Anthropic is not documented, so ask for the reference in the case.

Does status.claude.com cover Bedrock and Google Cloud?

No. The page lists claude.ai, Claude Console, Claude API, Claude Code, Claude Cowork and Claude for Government. Use the AWS Health Dashboard or Google Personalized Service Health for the cloud routes.

Can Team plan members contact Anthropic support directly?

Only through Fin, the AI bot. Human support on Team goes through the Primary Owner, Owners or Console Admins. Enterprise lets Owners designate additional support contacts, including entire groups.

Sources

1. Claude Help Center: How to get support (1 October 2026)
2. Claude Help Center: Support for Claude in Amazon Bedrock (1 October 2026)
3. Claude Help Center: Designate support contacts (1 October 2026)
4. Claude status page (1 October 2026)
5. Claude pricing (1 October 2026)
6. Claude Platform Docs: Service tiers (1 October 2026)
7. AWS Support plans (1 October 2026)
8. AWS Support pricing (1 October 2026)
9. AWS Support FAQs (1 October 2026)
10. AWS Service Terms, section 50.12 (1 October 2026)
11. Amazon Bedrock: Scaling and throughput best practices (1 October 2026)
12. AWS Support API: CreateCase (1 October 2026)
13. What is AWS Health (1 October 2026)
14. Google Cloud Customer Care pricing (1 October 2026)
15. Google Cloud: Anthropic's Claude on Google Cloud (1 October 2026)
16. Google Cloud Service Specific Terms (1 October 2026)
17. Google Cloud: Personalized Service Health overview (1 October 2026)
18. Microsoft Learn: Claude hosting options in Foundry (1 October 2026)
19. Azure support plans (1 October 2026)