Documentation
Everything you need to build with Arble. Installation, building, automation, deployment, and operating it in production.
Every topic is its own page. Use the sidebar to browse by section, or search to jump straight to a command, an endpoint or an error code.
Quick start
Four steps from nothing to a working agent. Do them in order — each assumes the previous one.
Install
Install the desktop app, the CLI, or both, on macOS, Linux or Windows.
3 min read
Pair Device
Link a phone or second machine so sessions and device control span both.
4 min read
Run your first agent
Model selection, the first permission prompts, and where state is written.
5 min read
Connect an MCP server
Connect any Model Context Protocol server; its tools join the same registry.
6 min read
Browse documentation
Getting Started
Setup and first contact. Nothing here assumes prior knowledge of agent systems — read it once, then rarely again.
6 pages
Agent
The parts of Arble you interact with daily. Each concept here has a direct equivalent in the CLI and the API, so learning it once covers all three.
7 pages
Developers
The extension surface. Five entry points, complementary rather than alternative — most real integrations use two or three.
5 pages
Reference
Lookup material. You arrive here from search, not from a table of contents.
4 pages
Popular guides
- Install — Install the desktop app, the CLI, or both, on macOS, Linux or Windows.
- Memory — Durable per-project facts, retrieved semantically and outliving sessions.
- Tool SDK — Build native tools in Python, TypeScript, Go or Rust.
- CLI — Drive every part of Arble from the terminal.
- API Reference — REST and streaming API: sessions, runs, memory, tools, permissions.
- Permissions — The gate every tool call passes, whatever initiated it.
Latest updates
| Page | Change |
|---|---|
| Tool SDK | New — decorators, schemas, streaming, publishing |
| CLI | New — full command reference, scripting, CI |
| API Reference | New — REST, SSE, webhooks, errors |
| Self-hosting | New — Docker, Kubernetes, HA, DR |
| Memory | Expanded — encryption, reindexing, retention |
Playbook
The Playbook is not product documentation. It teaches the concepts underneath AI systems — context windows, planning, tool calling — using Arble as an example rather than a subject. Read it to understand why an agent behaves as it does; read these guides to know what to type.