Skip to content

docs: login routes and European data handling in engine authentication - #172

Merged
makeitfutureDev merged 3 commits into
betafrom
docs/engine-authentication-europe
Oct 14, 2026
Merged

makeitfutureDev merged 3 commits into
betafrom
docs/engine-authentication-europe

Conversation

@makeitfutureDev

Copy link
Copy Markdown
Owner

Design note

The public docs page configuration/engine-authentication explained how to sign in Claude and Codex but not what each login means for a European customer. This change keeps the existing sections and adds:

  • Login routes at a glance: one table of every shipped engine login and where it is configured (Claude claude login / setup-token / ANTHROPIC_API_KEY; Codex ChatGPT sign-in / API key, gateway or per-conversation; Qwen and Qwen EU provider keys; OpenCode operator setup, read-only and network-off).
  • Choose a login for work in Europe: a separate section with two parts. Internal work without customer data covers personal Claude Pro/Max and ChatGPT Plus/Pro logins (consumer terms, no DPA, Anthropic EEA terms limit to non-commercial use). Customer data under a processor contract is a table of the commercial routes: Claude Team/Enterprise seat or API key (Commercial Terms + DPA, global/us inference only, US storage), ChatGPT Enterprise and Business sign-in, OpenAI API key, Qwen EU harness (Frankfurt, EU deployment scope), and Claude via Amazon Bedrock / Google Vertex AI.

Each row carries a Status in ChannelGate column verified against the source on beta:

  • Bedrock/Vertex: not a login route. src/gateway/claude-login.js resolves only setup token, operator login, engine-home login or ANTHROPIC_API_KEY; src/engines/child-env.js passes no CLAUDE_CODE_USE_*, AWS_* or Google variables.
  • OpenAI EU API endpoint: not reachable. CODEX_API_RELAY_RULE in src/gateway/egress/catalog-rules.js swaps the API-key placeholder only on api.openai.com.
  • Qwen EU: the shipped EU-resident route (src/engines/qwen.js, qwen-eu, Alibaba Cloud Model Studio eu-central-1).

Provider facts cite the providers' own pages (Anthropic data residency and legal terms, Claude Code Bedrock/Vertex docs, OpenAI data controls and business-data pages, Alibaba Model Studio regions) as of October 2026. The section says plainly that it is orientation for the reader's own assessment, not legal advice, and avoids any compliance claim about ChannelGate itself.

Also fixes an inherited en dash ("Claude–Codex failover") in the same page and updates the page description in src/data/handbook.mjs.

Checks

documentation$ npm run build
[build] 152 page(s) built in 3.25s
Verified 45 functionality areas and 108 gateway + 10 library controls, each documented exactly once.
Verified 151 Starlight pages and 26834 internal links, with the search index.

documentation$ npm test
Verified export preserves website, blog, API routes; merges docs sitemap once; validates before writes; preserves literal Markdown.

External links: all 11 resolve; openai.com/business-data and the help.openai.com residency article return 403 to a bare curl because of bot protection but open in a browser.

🤖 Generated with Claude Code

makeitfutureDev and others added 3 commits October 14, 2026 20:10
…ation

Add a "Login routes at a glance" table covering every shipped engine login
(Claude login, setup token, API key; Codex ChatGPT sign-in and API key;
Qwen and Qwen EU provider keys; the OpenCode read-only setup), and a new
"Choose a login for work in Europe" section that separates consumer logins
for internal, non-customer material from the commercial routes that carry a
data processing addendum.

The Europe table states what ships today: Claude Team/Enterprise and API-key
logins are supported with US or global inference and no EU option on any
first-party Anthropic plan; ChatGPT Enterprise/Business sign-in and the
OpenAI API key are supported with the residency caveats OpenAI documents;
the Qwen EU harness is the EU-resident route that ships (Alibaba Cloud Model
Studio, Frankfurt); the OpenAI European API endpoint is not reachable through
the Codex API-key relay, which is pinned to api.openai.com; and Claude via
Amazon Bedrock or Google Vertex AI is not a ChannelGate login route because
the Claude credential resolver and child environment allowlist do not carry
those variables.

Provider facts are cited to the providers' own pages as of October 2026.
The handbook description for the page is updated to match.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Signed-off-by: makeitfutureDev <144107855+makeitfutureDev@users.noreply.github.com>
…op Composio row

Per the product owner, Claude through Amazon Bedrock, Claude through
Google Cloud Vertex AI and the OpenAI Europe API endpoint are being
implemented, so the Europe table now lists them as routes marked
"In development" instead of describing them as unsupported. Adds the
AWS and Google Cloud source links for region and no-training facts.
Removes the Composio row from the Codex credential table; connected
app accounts are out of scope for an engine authentication page.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Signed-off-by: makeitfutureDev <144107855+makeitfutureDev@users.noreply.github.com>
State that the Europe section lists provider options only, that the DPA,
retention, training and residency commitments sit in the customer's own
contract with the AI provider, and that Makeitfuture and ChannelGate accept
no responsibility for provider data handling or model output and actions.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Signed-off-by: makeitfutureDev <144107855+makeitfutureDev@users.noreply.github.com>
@makeitfutureDev
makeitfutureDev merged commit 51a7042 into beta Oct 14, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant