Developers

Build agents without giving models root authority

MAPLE gives developers packageable agents, model-neutral inference, explicit capabilities, commitment gating, and replayable provenance without inventing those control surfaces from scratch.

Fastest Path

Clone the repo and inspect the Maple AI OS path

git clone https://github.com/mapleaiorg/maple.git
cd maple
cargo build -p maple-cli -p palm-daemon -p palm
cargo run --manifest-path examples/mwl-worldline-lifecycle/Cargo.toml
cargo run -p maple-cli -- daemon start --foreground

Install and quickstart

Use the docs for a clean install path and a 5-minute first run.

Open getting started

Author your first package

Create a Maplefile that matches the implemented schema and connects cleanly to OS policy.

Open tutorial

Use the API and integrations

Reference the REST surface, current CLI flows, and the workspace-first Rust plus HTTP integration paths.

Open API docs

Core Semantics

Three rules to design around

Cognition is advisory

Model output drafts intent. It does not grant execution authority.

Capabilities are explicit

Tool access begins at zero and is granted narrowly through Guard and package contracts.

Receipts matter

Every meaningful effect should remain inspectable through provenance history.

Need implementation support for your first governed workflow?

MapleAI can help package the first agent, design capability boundaries, and shape the production rollout path.

Talk to Developer Team Browse Docs