SubLaneSubLane
中文GitHub

LIGHTWEIGHT. SELF-HOSTED. TEAM-READY.

Your subscriptions.One shared lane.

A lightweight AI subscription gateway for internal teams. Self-host it on your infrastructure. Keep the essentials, skip the bloat.

Deploy SubLaneView on GitHub
ONE GATEWAY. END TO END.HTTP / SSE / WebSocket
Your subscriptions
CodexOpenAI
ClaudeAnthropic
AntigravityGoogle
SubLaneSubLaneSubLaneYour infrastructure
Account routingTeam permissions
Your tools
Codex CLI
Compatible clients

OpenAI · Claude · Gemini protocols

Credentials stay at the gateway. Members connect with their own keys.
One processGo + SQLiteYour infrastructureAGPL-3.0

Less to run.
More room to build.

For teams that prefer focused tools over bloated platforms. Connect accounts, control access, and understand usage in one small deployment.

A pool, with a plan.

Group subscription accounts, discover available models, and route requests to an eligible account.

  • Account groups
  • Model catalogs
  • Concurrency controls
Explore account routing

Personal keys. Team control.

Give each member their own gateway key. Keep subscription credentials separate and access scoped to the right group.

  • Member permissions
  • Key expiry
  • Model policies
See access controls

Know what happened.

Inspect requests, first-output latency, and reported token usage. Follow activity across members and models.

  • Request diagnostics
  • Usage summaries
  • Audit log
Understand team usage

Small by design.

A single Go process with an embedded interface and SQLite. No Redis or external database to operate. Deploy with Docker on your own infrastructure.

  • No Redis
  • No external database
  • Backup & restore
Read the deployment guide

From subscriptions to your team.

A short setup. A single place to manage it.

  1. 01

    Connect an account

    Authorize a Codex, Claude, or Antigravity subscription from the admin interface.

  2. 02

    Set the boundaries

    Create account groups and grant members access to the models they need.

  3. 03

    Bring your client

    Create a personal API key, set your gateway URL, and choose an available model.

Your server.
Your starting point.

Deploy SubLane with Docker. The installer picks a published release and creates your instance.

Manual deployment
Install with Dockerbash
curl -fsSL https://raw.githubusercontent.com/murongg/SubLane/main/scripts/install.sh | bash
Then open 127.0.0.1:8080 to create your administrator.

Requires Docker with Compose, curl, and jq.Review install script

A few things to know.

Which subscriptions can I connect?

Codex, Claude, and Antigravity subscription accounts, using browser authorization or credential JSON import. Official API-key upstreams are not supported yet.

Which clients and protocols are supported?

SubLane exposes OpenAI-compatible Responses and Chat Completions, native Claude Messages, and Gemini generation. HTTP/SSE and Responses WebSocket are implemented. Live account and desktop-client acceptance remains a separate verification step.

Does it include billing or token budgets?

No. SubLane provides request limits, reported usage, and Codex quota awareness. It does not implement billing or per-member token budgets. Claude and Antigravity quota readers are not available yet.

Where does my data live?

On your own infrastructure. SubLane stores application data in SQLite and encrypts subscription credentials with an instance-local key. Keep the database and credentials.key together when backing up.