Working with coding agents
Practical guides to the tasks, environments, and reviews around the code. Start with the problem your team is trying to solve.
Software factories
What is an AI software factory?
Understand AI software factories, where coding agents fit, and the practical questions to ask about context, execution, verification, and ownership.
Designing a software factory architecture
A practical software factory architecture covering task state, isolated execution, checks, retries, permissions, and the human acceptance decision.
Agent workflows
Coding agent orchestration for a software team
Organize coding agents around clear task boundaries, shared team context, independent execution, and review capacity. Includes a worked team workflow.
How to run parallel coding agents without losing the review
Run parallel coding agents with clear task boundaries, separate environments, and a review process that catches conflicts before they reach main.
Git worktrees vs cloud sandboxes for coding agents
Compare Git worktrees and cloud sandboxes for parallel coding agents, with practical examples for branches, port collisions, databases, and review.
How to verify coding agent work before it reaches users
Verify coding agent work with acceptance checks, focused tests, browser evidence, and a clear handoff that shows what passed and what remains uncertain.
Cloud coding agents: how remote work continues and recovers
Understand how cloud coding agents run away from your laptop, what can interrupt their work, and how to recover sessions and verify their results.
Shared team context for coding agents
Decide what belongs in repository rules, agent skills, MCP tools, and task handoffs, with a practical example for keeping team context current.
Coding agent costs: budget for accepted changes
Measure coding agent costs across subscriptions, model usage, environments, retries, and review. Build a budget around accepted changes with a worked example.
Taking a vibe-coded prototype into production
Turn a vibe-coded prototype into a repository your team can maintain, with reproducible setup, acceptance checks, and reviewable changes.
Integrations
Delegate work from Claude Code to Uno through MCP
Connect Claude Code to Uno over HTTP and OAuth, delegate a repository task, confirm execution, and follow the result without creating duplicate work.
From a Sentry issue to a reviewed pull request
Use a new Sentry issue to start a scoped investigation, reproduce the failure, and review a proposed fix with its checks in Uno.