Until recently, AWS-based organizations faced a trade-off: Amazon Bedrock kept Claude inside familiar AWS territory but lagged the first-party API on newer features, while going direct to Anthropic meant onboarding a new vendor and moving traffic outside the cloud relationship. Claude Platform on AWS is built to dissolve that trade-off. It is operated by Anthropic — same platform engineering as the first-party API — but it runs on AWS and authenticates with AWS-native mechanisms.
Who operates what
The distinction from Bedrock is about the operator, not the location. Bedrock is an AWS-operated managed service that offers Claude among other models; Claude Platform on AWS is Anthropic-operated infrastructure that happens to run on AWS. Both keep your traffic and identity story inside AWS. The operational difference shows up in one place above all: feature velocity.
Same-day parity, in practice
Claude Platform on AWS has same-day parity with the first-party Claude API: when Anthropic ships a new API capability, it is available here the same day it ships — with a handful of documented exceptions, mostly admin and billing features. As of July 2026 that means the full current feature set — the Batch API, Files API, code execution tool, web fetch tool, and web search tool, none of which are available on Bedrock — plus Managed Agents, which exist only on the first-party API and this platform.
For roadmap planning this changes the question you ask. On Bedrock or Vertex the question is "is the feature I need available, and if not, when?" Here the question collapses to "has Anthropic shipped it?" Teams building agentic products or batch-heavy pipelines feel this difference immediately; teams doing straightforward summarization may never notice it. Why Same-Day API Parity Matters works through the roadmap math.
Authentication: SigV4 plus a workspace ID
There are no API keys to issue or rotate. Requests are signed with AWS Signature Version 4 (SigV4) — the same request-signing scheme every AWS SDK call already uses — so your existing AWS credentials and IAM discipline carry over unchanged. The one new concept is the workspace: the SDK reads two environment variables, AWS_REGION and ANTHROPIC_AWS_WORKSPACE_ID, and the workspace ID tells the platform which of your organization's workspaces the request belongs to. The full mechanics are in Workspaces and SigV4.
import os
from anthropic import AnthropicAWS
os.environ["AWS_REGION"] = "us-east-1"
os.environ["ANTHROPIC_AWS_WORKSPACE_ID"] = "<your-workspace-id>"
client = AnthropicAWS()
message = client.messages.create(
model="claude-sonnet-5",
max_tokens=1024,
messages=[{"role": "user", "content": "Hello from our AWS account."}],
)
print(message.content)Note the model ID: bare claude-sonnet-5, exactly as on the first-party API. Only Bedrock uses the anthropic. prefix, which makes code written for this platform portable to and from Anthropic's own API with essentially no changes beyond the client class.
What to verify with your own teams
Because the platform is Anthropic-operated, your security and legal reviews should treat it as its own line item rather than assuming Bedrock conclusions transfer wholesale. The questions are the usual ones — data flow, retention, region behavior, audit surface — and the standard caveat applies: the platform inherits the compliance posture of the underlying cloud arrangement, and specifics should be confirmed with the providers rather than taken from any guide. On commercials, list pricing matches first-party list pricing, consistent with the other cloud surfaces.
Where to go next
Get a working call in minutes with Your First Call on Claude Platform on AWS, then go deeper on authentication in Workspaces and SigV4. To weigh this surface against Bedrock, Vertex, and Foundry, see the platform comparison.