Quickstart

Sleev is a context-management gateway for AI coding agents, bundled with a small command-line tool, sleev, that installs and manages it. Both run entirely on your own machine, so your data stays local and private. Sleev sits transparently between your harness and your model provider, and it never changes how you use your harness.

Harnesses resend their entire conversation on every turn, which burns tokens quickly as a session grows. Sleev intelligently manages that history by compressing the stale parts into compact summaries while keeping recent and relevant context and your provider's prompt cache intact, so you send far fewer tokens upstream and save a significant amount on usage.

Install

Start by installing the sleev command-line tool, which you use to set up and manage the gateway. On macOS and Linux, the easiest way is the install script, which downloads sleev and adds it to your path.

If you prefer npm, you can install sleev with npm instead, which works on macOS, Linux, and Windows:

Setup

With sleev installed, run it to open the TUI.

Sign in

The first time you run sleev, it asks you to sign in to your account. This links the gateway on your machine to your Sleev account so your usage and metrics show up in your dashboard.

Command-line setup

If you want to use only the CLI, run each step from the terminal.

Sign in to your Sleev account:

Register this machine and start the local gateway:

Check the gateway

After setup, confirm the gateway is running locally:

Stay up to date

Each time you open sleev, it checks for a newer release. If one is available, a notification appears in the corner with the new version and an Update button that upgrades the CLI and the gateway in one step.

Or update from the command line at any time:

If you want to pin a version, do so like this: