Build, publish, and install AI capabilities safely.

OlyWork is the capability packaging and secure install layer for Skills, MCP servers, Agents, and Workflows. These docs explain the product model, creator flow, buyer install path, integration surfaces, and launch operations.

bash

npm install -g @olywork/cli

Install the buyer-facing secure install CLI.

bash

oly install <token> --target cursor

Install an entitled package into a local runtime.

bash

oly mcp <capability> --setup claude-desktop --write

Generate and write an MCP client connector.

Start here

Quickstart

Publish and install the first OlyWork capability in the shortest path.

Capability packaging

Understand Skill, MCP, Agent, and Workflow package structures.

CLI reference

Install, sync, verify, update, rollback, and generate client configs.

Guides

Creator guide

Create packages, import source, set delivery mode, and submit for review.

GitHub import

Use GitHub as creator source while keeping buyer delivery protected.

MCP clients

Generate configs for Claude Desktop, Cursor, Windsurf, OpenClaw, and MCP clients.

API reference

Secure sync, integration config, runtime run, health, and webhook surfaces.

Monetization

Free secure install, paid secure install, protected bridge, and platform fee model.

Security model

Entitlements, artifact integrity, privacy boundaries, secrets, and safe logs.

Operate OlyWork

Architecture

How discovery, registry, entitlements, artifacts, CLI, and MCP gateway fit.

Launch runbook

Production env, storage, OAuth, smoke tests, launch gates, and go/no-go checks.

Troubleshooting

Recover install, publish, token, runtime, package, and GitHub import issues.

Core product surfaces

Secure install

Token-gated package sync with checksums and audit events.

GitHub import

Server-side repo ingestion for multi-file creator source.

MCP bridge

Client-specific MCP config for external developer runtimes.

Capability types

Separate schemas for Skills, MCPs, Agents, and Workflows.