---
title: Overview
summary: What Caveman is, which parts are open, and where to start reading.
canonical: https://docs.caveman.so/docs
updated: 2026-08-26T03:57:26+02:00
basis: inferred
---

# Overview

> What Caveman is, which parts are open, and where to start reading.
<DocSchema slug="" />

Caveman is an efficiency stack for AI agents. It sits between an agent and the model API it calls, and it
removes cost at four different points. Each point is a separate piece of software with its own licence,
and you can adopt any one of them without the others.

[Diagram: the four layers of the stack, bottom to top: Skill, Engine, Memory, Cloud.]

The layer at the bottom is a set of instructions. It has no runtime, no daemon, and no account. The layer
at the top runs on our machines and is the only paid part. This site documents everything below the
network boundary in full. For the managed plane it documents the shape, not the internals.

## What each layer does

### 00 Skill

A skill is a persistent instruction set your coding agent reads before it answers. It cuts filler,
narration and hedging out of the agent's own output, and it is explicit about what it must never touch:
code blocks, function names, CLI commands and exact error strings stay byte for byte.

It costs nothing to run because it is not software. It is text your agent already reads.

### 01 Engine

The engine compresses payloads on their way to a model. It detects what a payload is, routes it to a
compressor built for that shape, and stores the original bytes under a content-addressed handle before it
removes anything. If the agent needs what was dropped, it asks for the handle and gets the original back
exactly.

### 02 Memory

cavemem holds the context you would otherwise paste into every session. It stores text once, recalls the
parts a question actually needs, and compresses each hit through the engine on the way out.

### 03 Cloud

Caveman Cloud is the managed plane. It is where a fleet's numbers stop being local estimates and become
something a finance team can read. It is in private development and the waitlist is open.

## What the numbers mean

Every local tool on this site reports `inferred`. That word is load bearing. It means a per-run estimate
computed on your machine from an offline token counter, and it is never turned into a monthly figure or
a dollar amount.

No tool in this repository can emit `verified`. That label is reserved for hosted rollout systems that
can compare against a real bill.

<Note tone="honest" title="The short version">
If a number here has no basis label next to it, treat it as an example rather than a promise.
[How numbers are counted](/docs/counting) explains the four rules the code follows.
</Note>

## Where to go next

<CardGrid>
  <Card href="/docs/quickstart" title="Quickstart" meta="5 min">
    Install the skill, then compress something real and read the report.
  </Card>
  <Card href="/docs/architecture" title="Architecture">
    The path one request takes, and where the open-core line falls.
  </Card>
  <Card href="/docs/skill" title="The skill" meta="MIT">
    One command, thirty agents, no runtime.
  </Card>
  <Card href="/docs/engine" title="The engine" meta="BSL 1.1">
    Content-aware compression that keeps what an answer depends on.
  </Card>
</CardGrid>