Cycles · Sprint · Flow Metrics · CLI · MCP
Sprint-Driven AI Development with vem cycle
Learn how to organise your AI-assisted work into goal-driven cycles, track throughput with flow metrics, and give your agents real-time sprint context.
Why Cycles Matter for AI Teams
Individual tasks are great for tracking atomic work, but they say nothing about goals, deadlines, or priorities. vem cycles solve this by grouping related tasks into a time-boxed, goal-oriented container — exactly like a sprint, but built for async AI collaboration.
When a cycle is active, every agent that reads project context through vem automatically receives a 'Current Sprint' block: its goal, open tasks, and progress percentage. This alone dramatically reduces the number of times you need to re-explain priorities to your AI assistant.
The Cycle Lifecycle
A cycle moves through three states: draft → active → closed. You create it in draft so you can add tasks before kicking it off, activate it when the sprint begins, and close it when you want to capture the final retrospective.
Create and activate a cycle
# Create a new cycle with a goal and optional end date
vem cycle create "Q2 Auth Hardening" --end 2026-03-07
# List all cycles to find the ID
vem cycle list
# Activate it when your sprint starts
vem cycle start CYCLE-001
# Set it as the active focus (injects context into agents)
vem cycle focus CYCLE-001Assigning Tasks to a Cycle
Add tasks to a cycle when creating them, or update existing ones. The cycle flag accepts the cycle ID and immediately registers the task in the sprint backlog.
Assign tasks to a cycle
# Assign at creation time
vem task add "Implement MFA flow" --cycle CYCLE-001
# Or update an existing task
vem task update TASK-042 --cycle CYCLE-001Flow Metrics and Throughput
vem tracks when each task enters 'in progress' and when it reaches 'done', giving you real cycle time and throughput figures at the end of every sprint.
Run vem task flow to see WIP counts, average cycle time, and throughput for the current or any historical cycle. These numbers help you calibrate how many tasks to commit to in the next sprint.
View flow metrics
# Flow metrics for active cycle
vem task flow
# Flow metrics for a specific task
vem task flow TASK-042
# Close the cycle and capture retrospective data
vem cycle close CYCLE-001Agent Awareness via MCP
When your AI agent is connected through the vem MCP server, it can call get_cycle_context() to receive the active cycle goal, all tasks in the sprint, and current completion rate — without you needing to explain it every session.
Combine this with get_flow_metrics() and your agent can prioritise what to work on next based on actual throughput data, not guesswork.
- get_cycle_context() — active cycle goal, tasks, and progress
- get_flow_metrics() — WIP count, average cycle time, throughput
- vem cycle focus <id> — set which cycle agents see as active