selan.aiCompare · LiteLLM and Selan
Is there a LiteLLM alternative for a team's coding tools?
Yes, if what you need to govern is a team's Claude Code, Codex and Claude Desktop rather than your own application's model calls. LiteLLM is an open-source proxy you deploy yourself, with 100+ providers behind the OpenAI format. Selan is a hosted gateway: each developer signs in once, and every request is recorded and limited under their name. For an app of your own on many providers, LiteLLM is the better answer.
How do LiteLLM and Selan compare?
| LiteLLM | Selan | |
|---|---|---|
| What it is | A Python SDK and a proxy server, which it calls an AI gateway | A gateway, with a CLI that launches Claude Code and Codex through it |
| Where it runs | On your own infrastructure, from a Docker image or pip. Keys, the Admin UI and spend tracking need Postgres, and Redis once you run more than one instance | A hosted service, built and hosted in the EU. Nothing to run |
| Providers | 100+, in the OpenAI format | Anthropic, OpenAI, Google Vertex, OpenRouter, nexos.ai and Ollama, through the Anthropic Messages API and the OpenAI Responses API |
| A developer's credential | A virtual key, set as ANTHROPIC_AUTH_TOKEN | One sign-in with Google or Microsoft through selan login; no key on the laptop |
| Budgets | Per key, user, team, team member and end customer, reset over any duration, with rpm and tpm limits | Per person, by day, week or month in UTC, on pay-per-use keys; a key can carry a limit of its own |
| Secret redaction | Enterprise only | Built in, off until an owner turns it on |
| Who changed what | Audit logs are Enterprise only | Every switch and limit records who changed it and when |
| Cost | MIT licence outside its enterprise/ directory; Enterprise is priced by deployment size | Free for up to three seats, no card |
LiteLLM is a gateway you run for your applications. Selan is a gateway your team signs in to.
When is LiteLLM the better answer?
When you are building your own product on many models and want the gateway inside your own infrastructure. LiteLLM reaches far more providers than Selan does. Its code is open under the MIT licence, apart from the enterprise directory, so you can read and change it.
Its router balances load across deployments, retries, and falls back to another model group when a call fails. Budgets can belong to a key, a user, a team or one of your own customers, and rate limits sit on top.
If no team runs coding tools through it, you do not need Selan.
When is Selan the better answer?
When the traffic is a team's coding tools and nobody wants to run a gateway. A developer
runs selan login once, then selan claude or
selan codex. Claude Desktop goes through the same gateway with one settings
file.
The owner's Usage page ranks people and teams over 24 hours, 7 days or 30 days. Logs show every request with who ran it, the model and the price. There is no Postgres, Redis or proxy upgrade to look after.
The trade-off is that requests travel through Selan, so if Selan is unavailable those requests fail. Your accounts can still be used directly.
What the Usage page shows is in how to see Claude Code usage per developer, and the setup in the docs.
Can LiteLLM and Selan use Claude and ChatGPT subscriptions?
LiteLLM, from version 1.81.14, forwards each developer's own Claude subscription login to Anthropic. It prices that usage at Anthropic's API list prices and counts it toward budgets, though Anthropic bills it to the subscription. It does not see which Claude account a token belongs to.
For ChatGPT, a LiteLLM proxy uses exactly one login for every caller. Those models have no price in its cost map, so recorded spend is $0 and budget limits never trigger.
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. Selan reads each Claude account's 5-hour and 7-day windows from every response. It prices a seat as what the same requests would have cost per token, next to what your keys actually cost.
How do spend limits and secret scanning differ?
LiteLLM's budgets are the more flexible of the two, with any reset duration and rpm and tpm limits. Its secret detection and redaction needs an Enterprise licence. So do audit logs and single sign-on for more than five users.
A Selan limit applies across every pay-per-use key the company has added, so switching keys does not start it again. It binds the moment it is reached, and the refusal says what stopped it and when it lifts. With secret scanning on, a secret it recognises is replaced with a placeholder before the request leaves for the provider.
Limits are on the spend page, and scanning in the secret scanning docs.
How do you move Claude Code from LiteLLM to Selan?
Behind LiteLLM, Claude Code reads ANTHROPIC_BASE_URL pointing at the proxy and
ANTHROPIC_AUTH_TOKEN holding a virtual key. selan claude sets
both for the session it launches, in place of whatever the shell exported.
An owner connects the company's provider accounts and keys in Selan and switches on the models the team may use. Your own applications can stay on LiteLLM, or call Selan's API with a base URL and a token.
Free for up to three seats, and the Free plan needs no card.
The short version
- LiteLLM is open source, runs on your infrastructure and reaches 100+ providers.
- Selan is hosted in the EU and records Claude Code, Codex and Claude Desktop per person.
- LiteLLM's secret redaction, audit logs and SSO beyond five users need an Enterprise licence.
- You can run both: LiteLLM for your own apps, Selan for the team's coding tools.
What LiteLLM does is as its own README, licence and documentation at docs.litellm.ai described it at version 1.105.1, read on 14 October 2026. What Selan does is selan-cli and selan-control as they stood on the same day.