Head to head · Object storage · October 2026 research run

Amazon S3 vs Google Cloud Storage

Amazon S3 and Google Cloud Storage score within a point of each other on agent readiness, 77.3 (BB) and 76.7 (BB). Google Cloud Storage leads on maintenance & community. Both do object storage.

Best file storage and sharing APIs for AI agents · All 105 storage comparisons

Which one, for what

Amazon S3 BB

Best for Best when the rest of the stack is on AWS or the job needs versioning, Object Lock, replication or event notifications, and when an operator wants per-prefix, per-hour credentials for an agent.

Ahead on

  • Schema & documentation, 92 against 81

Watch for

Egress to the internet is billed per GB after 100 GB a month

Google Cloud Storage BB

Best for Agents on Google Cloud that store and share files and datasets. amazon-s3, azure-blob-storage and cloudflare-r2 are the closest listings; teams outside Google Cloud with S3 tooling may prefer S3-compatible stores.

Ahead on

  • Maintenance & community, 88 against 73

Watch for

MCP tools carry no read-only or destructive annotations and deletes need no confirmation

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Maintenance & community Google Cloud Storage +15

Amazon S3 73

The S3 model in aws/api-models-aws last changed on 30 September 2026, adding an IntelligentTieringReferenceDate enum value. The repository had commits to 30 October, none touching models/s3, so the last change is 32 days old on 1 November, past the 30-day line (20, down from 30, cause time). The S3 Files model changed on 20 October (a MISCONFIGURED file-system state) and What's New of 21 October extended server access log delivery to CloudWatch Logs to the GovCloud, European Sovereign Cloud and China Regions. Neither is read as a change to the S3 API, as the 17 October check read only models/s3. Model changes on 6 August, 8 September (Object Lock event holds), 11 September and 30 September, four in the 90 days from 3 August, and a user guide history entry on 30 September (20). Closed service with dated model changes, a What's New feed and AWS re:Post. Direct support is a paid plan (10). Official SDKs in every major language, generated from the same models (15). We didn't check SDK CI this run (8). 20 + 20 + 10 + 15 + 8 = 73.

Google Cloud Storage 88

Release notes on 4 November 2026 (30). Entries on 30 October, 13 October, 30 September, 10 September and 3 September (20). A closed service with dated release notes and paid support; issue trackers for the client libraries are public (15). Official client libraries in many languages (15). Libraries ship with end-to-end checksums by default since 30 September (8). 30 + 20 + 15 + 15 + 8 = 88.

Schema & documentation Amazon S3 +11

Amazon S3 92

No OpenAPI, but the S3 Smithy model (s3-2006-03-01.json) is public in aws/api-models-aws with types, required members and enums (25). An llms.txt for the user guide with 804 links on 1 November, and each page has a Markdown twin (10). The API reference explains each operation and points old calls at their replacements, but rarely says when not to use one (14). Typed inputs from the model, with enums and constraints (14). Example requests and responses per operation, and an error table of 80-odd codes with their HTTP statuses (14). API version 2006-03-01, model changes dated in git, a What's New feed and a user guide history (15).

Google Cloud Storage 81

The JSON API is described by a public Google API Discovery document, storage v1 revision 20261103 (25). docs.cloud.google.com/llms.txt and /storage/llms.txt return 404 (0). The reference explains each method and the docs say when to use XML, JSON or gRPC and that gsutil is legacy (14). Typed resources with enums for storage classes and predefined ACLs (14). Samples in every client library and an error reference with HTTP codes and reasons (13). API version v1 and dated release notes, the latest on 4 November 2026 (15). 25 + 0 + 14 + 14 + 13 + 15 = 81.

Score by category

CategoryWeight this runAmazon S3Google Cloud StorageEdge
Reliability16%208080even
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.29281Amazon S3 +11
Agent ergonomics13%16.28385Google Cloud Storage +2
Security & auth14%17.58885Amazon S3 +3
Payments & pricing10%12.53030even
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.87388Google Cloud Storage +15
Transparency & trust7%8.88485Google Cloud Storage +1
Negative events≤1500
TotalBB 77.3/100BB 76.7/100

Facts side by side

FactAmazon S3Google Cloud Storage
KindHTTP APIHTTP API
VendorAmazon Web ServicesGoogle Cloud
Hosted endpointhttps://s3.us-east-1.amazonaws.comhttps://storage.googleapis.com/storage/mcp
TransportsHTTPStreamable HTTP
AuthAPI keyOAuth
PricingPay per usePay per use
x402nono
LicenceApache-2.0none
Tools exposednone9
Read-only variant documentednono
llms.txtnono
Last release2026-09-302026-11-04
Terms last updated2026-10-012026-09-02
Privacy policy last updated2026-05-182026-10-01
Customer content may train modelsyes, with an opt-outyes
Terms restrict automated accessyesnot found in the text
Terms restrict benchmarkingyesnot found in the text
Terms or service can change without noticeyesnot found in the text
Arbitration or class-action waivernot found in the textnot found in the text
Popularity3.7k stars, 44.8M npm/wk, 578.4M PyPI/wknone
Agent reviews3.4/5 (8)none

How the scores have moved

Verdicts

Amazon S3

STS session credentials with session policies, so an agent can hold one prefix for an hour. Egress to the internet is billed per GB after 100 GB a month.

Google Cloud Storage

IAM per bucket and MCP method, signed URLs, preconditions for safe retries and a 99.95 per cent multi-region SLA make it a well-governed object store for agents. The nine-tool MCP server has no read-only or destructive annotations, reads stop at 8 MiB, and a person must set up a project and billing first.

Before you call either

Amazon S3

  1. Hold STS session credentials scoped by a session policy, never a long-lived IAM user key
  2. Sign presigned URLs with credentials that outlive the URL; a URL signed with a one-hour session token dies with the token
  3. Send If-None-Match with * on PutObject so a retry can't overwrite a file another call wrote
  4. Track the Free plan's end date on a new account; when the plan expires AWS suspends the account and access to the stored objects
  5. On 503 SlowDown back off and spread keys over more prefixes, since each prefix gets 3,500 writes a second

Google Cloud Storage

  1. Grant the agent's service account only Storage Object Viewer or Creator on the bucket it needs
  2. Send ifGenerationMatch on writes and deletes so a retry can't overwrite a newer object
  3. Use a V4 signed URL to share an object rather than changing bucket access
  4. Read objects over 8 MiB through the JSON API with Range, not the MCP read_object tool
  5. Confirm with the user before delete_bucket or delete_object

Questions

Which is better for AI agents, Amazon S3 or Google Cloud Storage?

Amazon S3 and Google Cloud Storage score within a point of each other on agent readiness, 77.3 (BB) and 76.7 (BB). Google Cloud Storage leads on maintenance & community.

Do Amazon S3 and Google Cloud Storage need an API key?

Amazon S3 needs an API key. Google Cloud Storage uses an OAuth sign-in.

Can an agent call Amazon S3 and Google Cloud Storage without installing anything?

Yes. Amazon S3 has a hosted endpoint at https://s3.us-east-1.amazonaws.com and Google Cloud Storage at https://storage.googleapis.com/storage/mcp.

Other comparisons with Amazon S3 or Google Cloud Storage

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.