An autonomous AI team that runs
full Scrum sprints while you sleep

AiScrum Pro orchestrates GitHub Copilot CLI to execute complete sprint cycles — refinement, planning, parallel execution, code review, and retrospectives — with real quality gates, drift control, and escalation boundaries.

↓ scroll
Built on AI-Scrum

The engine behind the framework

The AI-Scrum Framework defines how humans and AI collaborate — with a manifesto, ceremonies, boundaries, and escalation rules.

AiScrum Pro is the runtime that executes it. It connects to GitHub Copilot via the Agent Client Protocol and runs sprints autonomously — creating branches, writing code, running tests, opening PRs, tracking velocity, and improving its own process with each retro.

You are the Stakeholder. You set direction, drop ideas, approve deliverables. The AI team handles everything else.

👤
Stakeholder
Sets direction
Reviews work
Has veto
direction → ← deliverables ← escalations decisions →
🤖
AI Team
Lead Agent
Worker Agents
Challenger Agent
Sprint Lifecycle

Five ceremonies, fully automated

Start a sprint and come back to finished, tested, reviewed code.

01
🔍

Refine

Ideas decomposed into concrete issues with acceptance criteria

02
📋

Plan

ICE scoring, scope selection, milestone assignment, dependency graph

03

Execute

Parallel workers in isolated git worktrees, each quality-gated

04
📊

Review

Sprint metrics, velocity tracking, deliverable summary for stakeholder

05
🔄

Retro

Process improvements applied to agent prompts and workflows

Capabilities

Not a chatbot wrapper

A structured sprint engine with real project management, quality enforcement, and process improvement.

🔀

Parallel Execution

Multiple issues worked simultaneously via isolated git worktrees with automatic merge pipelines.

🛡️

7 Quality Gates

Tests exist, tests pass, lint clean, types clean, build passes, scope check, diff size — enforced on every issue.

📈

Velocity Tracking

Sprint metrics, issue completion rates, cycle time analysis. Every sprint gets smarter.

🌐

Web Dashboard

Real-time sprint control center with 9 views, live agent chat, and historical sprint navigation.

💬

Agent Chat

Open ad-hoc ACP sessions with pre-configured roles — researcher, planner, reviewer, challenger.

🔔

Push Notifications

ntfy.sh integration for real-time alerts when sprints complete, issues block, or decisions are needed.

🧪

Test Isolation

Run test sprints with separate prefix, milestones, and branches — fully isolated from production.

📝

Sprint Documentation

Automatic sprint logs, huddle notes, velocity reports, and ADR maintenance.

⚙️

Config-Driven

One Zod-validated YAML file controls the entire engine — models, gates, parallelism, merge strategy.

Why AiScrum Pro

Structure beats ad-hoc

Ad-hoc AI Coding AiScrum Pro ✦
PlanningNone — chat until it worksICE-scored sprint backlog, dependency graphs
ExecutionOne issue, manually drivenParallel workers via git worktrees
Quality"It should work"7 enforced gates per issue
MemoryLost every sessionSprint logs, velocity, issue comments, ADRs
Scope controlFeature chasingDrift detection, sprint lock, escalation
ImprovementStatic promptsRetro-driven process evolution
Dashboard

Your sprint control center

Monitor progress, chat with agents, navigate sprint history — all in one place.

Product Backlog
Product Backlog
Sprint Report
Sprint Report
Blocked Issues
Blocked Issues
Decisions Pending
Decisions Pending
Ideas Inbox
Ideas Inbox
Settings
Settings
Logs
Logs
Sprint Backlog
Sprint Backlog
Controls

Autonomy with accountability

Autonomous execution needs boundaries. Every one of these is built in and enforced.

🔒

Drift Control

Sprint scope is locked after planning. Discovered work goes to backlog, never into the current sprint. If >2 unplanned issues appear, the engine escalates immediately.

⚖️

Escalation Model

The AI decides how, never what. Strategic direction changes, ADR modifications, scope changes, and dependency additions always require stakeholder approval.

🏛️

Challenger Agent

An adversarial reviewer that challenges assumptions, finds blind spots, and prevents groupthink before sprint review. The team's built-in devil's advocate.

📋

Definition of Done

Acceptance criteria before coding. Tests that verify behavior. PR reviewed. CI green. Issue closed with summary. No shortcuts, no exceptions.

Foundation

Built on the AI-Scrum Framework

AiScrum Pro implements the AI-Scrum Framework — an open-source methodology for human-AI software development built on real Scrum principles.

The framework defines the roles, ceremonies, boundaries, and values. AiScrum Pro is the TypeScript engine that executes them via the Agent Client Protocol.

Read the full framework →
AI-Scrum Manifesto
  • Autonomous execution over constant approval
  • Verified evidence over claimed completion
  • Sprint discipline over feature chasing
  • Continuous process improvement over static workflows

"While there is value in the items on the right, we value the items on the left more."

Get Started

Up and running in 60 seconds

~/my-project
# Pick a starter config for your stack
$ cp -r examples/python/.aiscrum .aiscrum
# Install and launch
$ npm install && npx tsx src/index.ts web
🏃 AiScrum Pro dashboard → http://localhost:9100
# Or run a full sprint cycle
$ npx tsx src/index.ts full-cycle --sprint 1
🔍 Refining ideas... → 📋 Planning sprint... → ⚡ Executing...
# Starter configs: typescript · python · react · go

Let the AI team handle the sprint

Open source. TypeScript. Config-driven. Your code, your rules, your repo.