Getting Started

5-Minute Quickstart

Clone the repo, boot the demo, and understand the commitment flow.

Prompt pack
M-05b
Source material
  • maple/docs/getting-started.md
  • maple/docs/tutorials/worldline-quickstart.md

5-Minute Quickstart

This is the fastest path from zero to seeing MAPLE act like an Agent OS instead of a prompt wrapper. The demo boots the kernel, creates worldlines, sends an action through commitment gating, records provenance, and shows that the runtime treats consequence as a governed event.

Prerequisites

  • Rust 1.80 or newer
  • Ollama recommended, but optional

Run the demo

git clone https://github.com/mapleaiorg/maple.git
cd maple

# Optional local model runtime
ollama serve &
ollama pull llama3.2

# Run the MAPLE demo
cargo run -p maple-demo

What you are seeing

  1. Kernel boot: Event Fabric, Commitment Gate, and WorldLine ledger services initialize.
  2. Worldline creation: the demo registers example identities so the runtime has durable actors instead of ephemeral chat sessions.
  3. Coupling: one worldline establishes a relationship with another so signals can become meaningful interaction rather than raw transport.
  4. Meaning and intent: the runtime forms an interpretable plan before any side effect is allowed.
  5. Commitment gate: the proposed action is evaluated through policy, capability, and invariant checks.
  6. Consequence: only an authorized action crosses into execution.
  7. Provenance: receipts and outcomes are written so the entire sequence can be replayed or audited later.
  8. Invariants: the demo closes by validating the architectural guarantees that prevent silent unsafe execution.

If Ollama is not running

MAPLE should still let you explore the non-model parts of the flow: worldlines, commitments, receipts, and operator surfaces. That is intentional. The runtime is built so governance and provenance stay visible even when a local model backend is unavailable.

Next steps