Skip to content

Docs

For people, and for the agents helping them.

Every page here is also fetchable as plain markdown at the same path with /raw appended (or ending in .md), with no sign-in and no API key. See /llms.txt for the machine-readable index, or /llms-full.txt for the whole corpus in one fetch.

Start here

Start

New here. Install, and understand how Saylek works.

  • Quickstart

    Install the CLI, ask your first question, and recover when a step fails.

    New member

  • Concepts

    How sharing works between two people, what a shared machine exposes, receipts, and why a request can run elsewhere.

    New member · Evaluating Saylek · Members

Guides

What you can do with Saylek, one goal at a time.

  • Use the models you already serve

    Connect Ollama, LM Studio, llama.cpp or vLLM, check your models, and send a first request from an app with one API key.

    Machine owners · Developers

  • Start without a runtime

    Download one starter model with your consent, serve it on your GPU or Mac, and send a first request from an app.

    Machine owners · Developers

  • Make more of one GPU

    Serve several models from one GPU, know how they load and unload, keep your own work first, and limit what shared requests can use.

    Machine owners

  • Give important applications priority

    Give higher-priority applications precedence when requests are waiting for capacity.

    Developers · Members

  • Add machines and GPUs

    Connect more machines, choose which GPUs each one uses, and see how Saylek spreads requests across them.

    Machine owners

  • Share access, or use someone's

    Invite someone to use your machine, accept an invitation to use theirs, and pause or end access either way.

    Members · Machine owners

  • Set up Saylek for a team

    Run your team's machines from one account, or let each person keep their own, and choose who manages access and application priorities.

    Members · Evaluating Saylek

Sharing

Who can reach the models on your machine, and whose you can reach.

  • Sharing your models

    Let the people you choose reach the models your machine runs, and the three different ways to stop.

    Machine owners · Members

  • Sharing and access

    How access starts and ends between two people, and how to see who has it.

    Members · New member · Evaluating Saylek

  • Invite flow

    Invite someone to use the compute you share, and what happens when they accept.

    Members

Applications

Point an app, an SDK, or a coding agent at the models you can reach.

Troubleshooting

Start from the symptom you are seeing.

  • Troubleshooting

    Symptoms, what the state actually is, and how to get out of it.

    Members · New member · Developers · Machine owners

Reference

Generated tables and the command surface.

  • API reference

    Base URLs, supported endpoints, streaming, model ids, and how failure looks.

    Developers

  • Common commands

    The everyday command table and the leaving/pause/delete options, generated from README.md.

    Members · New member

Trust

What leaves your machine, and what we do not claim.

  • Privacy and egress

    What leaves your machine when somebody else serves you, and the two ways to stop it.

    Evaluating Saylek · Members · Developers

Terms, privacy, and security

  • Terms of Service: The beta terms, including the egress disclosure: a request your device cannot serve runs on another person's GPU.
  • Acceptable Use Policy: What may and may not be run on somebody else's machine, what running `saylek host wizard` commits you to, and who is answerable for what.
  • Privacy: What information Saylek handles, including prompts and completions, and what is retained.
  • Security: Who can use your machines, who can see your requests, passwordless sign-in, and how to report a security issue.