Skip to content

About

AI assisted delivery POC

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

65 Commits

Folders and files

Repository files navigation

Enterprise Delivery OS

License Test Contributions Welcome GitHub Stars GitHub Forks

v0.7.0 | Portable, role-based Delivery OS for host repositories

Delivery OS is a governed SDLC workflow layer for AI agents. Install it as .agency in a host repository to standardize roles, approvals, tracker integrations, and repeatable handoffs.

Key Value
Install location .agency inside the host repository
Best demo path Jira + Confluence with scm.provider="none"
Bootstrap ./.agency/bin/agency init --mode atlassian --docs atlassian

At A Glance

  • Role-based agents for Product, Planning, Developer, QA, Review, PM, Security, Architecture, and DevOps
  • Human approval gates built into the workflow
  • Portable integrations for Jira, Confluence, GitHub, Linear, and offline/fake modes

Contents

Requirements

  • Node.js 18+
  • Git
  • OpenCode CLI, install with npm install -g opencode

Quickstart

Run the following from the host repository root, not from inside the .agency checkout.

# Install OpenCode CLI
npm install -g opencode

# Add Delivery OS as a git submodule in the host repository
git submodule add https://github.com/klikaba/AI-assisted-delivery.git .agency
git submodule update --init --recursive

# Initialize the host repository
./.agency/bin/agency init --mode atlassian --docs atlassian

# For the Jira + Confluence demo path, set scm.provider to "none"
# in .agency-project.json before running OpenCode.

# Commit the tracked generated configuration
git add .agency-project.json .agency-rules.md .gitignore
git commit -m "chore: configure .agency"

# Start OpenCode with the generated config
opencode --config opencode.jsonc

If you want the guided bootstrap instead:

./.agency/setup.sh

If you get stuck:

./.agency/bin/agency doctor

Demo Path

The strongest current demo path is Jira + Confluence with scm.provider="none".

Recommended flow:

  1. Product refines a rough backlog item.
  2. Planning creates the implementation spec and execution plan.
  3. A human approves the spec in Confluence by changing Spec Status.
  4. PM syncs approval back to Jira labels.
  5. Dev, QA, Review, optional Security, and PM Release continue the governed workflow.

Primary workflow helpers:

  • workflow.product_refine
  • workflow.plan_finalize
  • workflow.dev_finalize

Use the detailed operator sequence in Delivery OS Demo Runbook.

Included Components

  • generated OpenCode config support
  • a local Delivery OS MCP server
  • provider adapters for Jira/Confluence, GitHub, Linear, and fake/offline mode
  • governed role prompts for the Delivery OS workflow
  • deterministic tests and simulated flows for regression protection

Workflow Roles

Core workflow roles:

  • Product Owner
  • Planning
  • Developer
  • QA
  • Code Reviewer
  • Project Manager

Conditional role:

  • Security

Optional specialist roles:

  • Architecture
  • DevOps

Reference

Get Help

For questions, issues, or support requests, email: support@klika.ba

Contribute

License

Apache 2.0 with attribution requirement.

If you use this software in a product or service, include the attribution below in your documentation or About section:

Built using the Klika AI Engineering Toolkit
https://github.com/klikaba/AI-assisted-delivery

See LICENSE and NOTICE.

About

AI assisted delivery POC

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages