Compare · Claude apps gateway and Selan
Is there an alternative to Anthropic's Claude apps gateway?
Yes: Selan is a hosted gateway for Claude Code, Codex and Claude Desktop, where Anthropic's
Claude apps gateway is one you run yourself. Anthropic's is free, ships in the
claude binary and keeps inference on your own cloud account. Selan needs no
server and also runs on Claude and ChatGPT subscriptions. If every request must stay in
your own cloud account, Anthropic's gateway is the better answer.
How do the Claude apps gateway and Selan compare?
| Claude apps gateway | Selan | |
|---|---|---|
| What it is | A server mode of the claude binary, started with claude gateway --config gateway.yaml | A hosted gateway, with a CLI that launches Claude Code and Codex through it |
| Where it runs | On your infrastructure: Linux, PostgreSQL 11 or later, HTTPS, and a hostname that resolves only to private addresses | A hosted service, built and hosted in the EU. Nothing to run |
| Clients | Claude Code and Claude Desktop | Claude Code, Codex and Claude Desktop |
| Where requests go | Amazon Bedrock, Claude Platform on AWS, Google Cloud's Agent Platform, Microsoft Foundry or the Anthropic API, with failover | Claude and ChatGPT subscriptions, Anthropic and OpenAI keys, Google Vertex, OpenRouter, nexos.ai and Ollama Cloud |
| Sign-in | Your OpenID Connect identity provider, such as Okta, Entra ID or Google Workspace. No SAML or LDAP | Google or Microsoft, through selan login |
| Policy | Model allowlists and managed settings per identity-provider group, in a YAML file. No admin UI | Models and limits switched in the Selan app, each change recorded with who made it and when |
| Usage | OpenTelemetry metrics with user, model, tokens and latency, sent to a collector you run | A Usage page ranking people and teams, and a log of every request with who ran it, the model and the price |
| CI jobs | No service-token flow; a pipeline calls the provider directly | An API token an owner mints, carried as SELAN_TOKEN |
| Cost | No licence or per-seat fee; you pay for inference and the compute it runs on | Free for up to three seats, no card |
Anthropic's gateway is one you run inside your network. Selan is one your team signs in to.
When is the Claude apps gateway the better answer?
When inference has to run through your own cloud account, for example for data residency. Anthropic designed it for that case. Selan connects neither Amazon Bedrock nor Microsoft Foundry.
It is Anthropic's own, built into the claude binary and tested alongside each
Claude Code release. The gateway itself sends nothing to Anthropic unless the Anthropic API
is one of its upstreams. No other vendor sits between Claude Code and your provider.
It also fits a company that signs in through Okta or another OpenID Connect provider and wants each group's settings locked on every laptop. A developer cannot override what the group's policy locks.
Anthropic's own docs add one more case: without an own-cloud requirement, Claude Enterprise may fit better than either gateway.
When is Selan the better answer?
When nobody wants to run a gateway. There is no server, database, identity-provider client
or private network to set up. A developer runs selan login once, then
selan claude or selan codex, from any network.
When the team uses Codex or GPT models as well as Claude. Both go through the same gateway, into one usage record.
When jobs run unattended. An owner mints a token for the job, and it spends under an address of its own. The setup is in Running unattended.
The trade-off is that requests travel through Selan, so if Selan is unavailable those requests fail. Your accounts can still be used directly.
Can either gateway use Claude or ChatGPT subscriptions?
The Claude apps gateway forwards to cloud credentials or an Anthropic API key, and those bill per token. A Claude or ChatGPT subscription is not among its upstreams.
Selan runs on the Claude and ChatGPT accounts and API keys your company connects. Requests go to subscription accounts first and fall through to a key billed per token only when they are capped. Which accounts cover which models is in Models and providers, and the cost of each route in Max plan or API.
What does a developer have to set up?
With the Claude apps gateway, every machine runs Claude Code v2.1.195 or later. An admin
pushes forceLoginMethod and forceLoginGatewayUrl in managed
settings, and the developer runs /login through the browser.
Claude Code connects only to a gateway on a private address, so away from the office the developer needs the VPN. Claude Desktop gets its own managed key and signs in separately.
With Selan, Claude Desktop takes one settings file, a relaunch and a sign-in; the steps are in Claude Desktop. No provider key reaches a laptop in either case.
How do their spend limits differ?
They are close. Both cap a person by day, week or month and reset on UTC calendar boundaries. The Claude apps gateway also caps a group or the whole organisation as a default for each member, set through an admin API.
Its spend is an estimate in US dollars, a circuit breaker rather than an invoice. If its database is down, enforcement lets requests through by default.
A Selan limit applies across every pay-per-use key the company has added, and binds the moment it is reached. Both refusals name when the limit lifts. Limits are on the spend page.
Free for up to three seats, and the Free plan needs no card.
The short version
- The Claude apps gateway is free, ships in the
claudebinary and runs on your infrastructure. - It forwards to Bedrock, Claude Platform on AWS, Google Cloud, Microsoft Foundry or the Anthropic API.
- Selan is hosted in the EU and adds Codex and Claude and ChatGPT subscriptions.
- If every request must stay in your own cloud account, choose Anthropic's gateway.
What the Claude apps gateway does is as Anthropic's Claude Code docs (Claude apps gateway, its configuration reference, deployment guide and spend limits) described it on 16 October 2026. What Selan does is selan-cli, selan-control and this site's docs as they stood on the same day.