The Agentic Dev Loop

Everything you need to design, build, and ship an internal dev system where humans and agents work side by side.

Share this with your team

The shift

From doing the work to watching the work get done.

01

How do I do the work?

Traditional development. You're the one writing code, running tests, debugging. The entire loop runs through your hands.

The developer era

02

How do I watch myself work?

AI-assisted development. You're still typing, but now you're observing suggestions, accepting completions, guiding the AI. Half doing, half watching.

The copilot era

03

How do I watch the agents work?

Agentic development. You're directing and reviewing, not writing. Agents execute. You orchestrate, verify, and course-correct.

The agentic era ← you are here

This shift frees you from typing to focus on product thinking, architecture, and decisions that matter. But it only works if you have the infrastructure to trust what the agents produce.

That's what this site is about: building the systems that let you confidently watch agents do the work.

Your team is integrating agents into development workflows. You need answers — about environments, identity, CI, review, governance, and audit — and you need them now.

This is the reference. It covers how to design:

  • execution environments
  • identity and secrets
  • CI pipelines
  • code review systems
  • governance controls
  • audit trails

…when agents are allowed to participate in the development loop.

In the AI era, inner loop speed and execution velocity matter more than ever. When agents become first-class participants, the loop itself must evolve.

Core Thesis

The problem is not building agents.

The real problem is designing the development system when agents become first-class participants:

Where code runs
Who is allowed to write code
How secrets flow
How PRs are created
How review happens
How trust is enforced
How every action is audited

In the agent era, the development environment is no longer a laptop. It is a distributed system composed of humans, agents, ephemeral runtimes, CI pipelines, reviewers, policy engines, and audit systems.

23 Guides

Technical pillars of the loop

View all guides
01

Execution Environments

Where do humans run code? Where do agents run code? When is execution local vs remote?

Read more
02

Ephemeral Runtimes

Should every task run in a fresh environment? How long do workspaces live?

Read more
03

Reproducible Toolchains

How do we guarantee identical behavior across runs? How do we eliminate snowflake laptops?

Read more
04

Identity, Secrets & Trust

Who is an agent, cryptographically? What credentials can it hold? How long do identities live?

Read more
05

Task Routing & Orchestration

Which agent handles which task? How is concurrency controlled? How are failures retried?

Read more
06

Who Writes Code?

When are agents allowed to author code? How is authorship recorded? Who owns the result?

Read more
07

Who Opens Pull Requests?

Can agents open PRs directly? How are branches named and scoped? How are tasks linked to specs?

Read more
08

Code Review

Which reviews are automated? What do humans still review? How is drift prevented?

Read more
09

Policy & Guardrails

What actions are forbidden? What violations are blocking? How is drift detected early?

Read more
10

Human-in-the-Loop

Where must humans approve? How are confidence thresholds set? How are overrides handled?

Read more
11

Observability

What do we log? How do we trace actions across systems? How do we measure quality and cost?

Read more
12

Audit Trails & Provenance

Who did what? In which environment? With which credentials?

Read more
13

Writing Specs & Prompts

How do you write specs agents can execute against? When is a prompt enough vs. a structured charter?

Read more
14

Local-First Dev Environments

How do you set up a local environment that agents can also use? What does the inner loop look like?

Read more
15

Handing Off Environments to Agents

How do you swap your running dev environment with an agent? How does context transfer work?

Read more
16

Ingress & Public Access

How do agents access running services? How do you expose local or ephemeral environments to the outside?

Read more
17

Agent-in-the-Loop Implementation

How does an agent participate during the implementation phase? What does the human-agent handshake look like?

Read more
18

The Agentic Build Loop

How do you use agents to build agents? What does the meta-loop look like?

Read more
19

Reproduce Steps

Before any implementation begins, how do you verify the bug or behavior you are trying to fix actually exists?

Read more
20

Expect & Demo Steps

After implementation, how do you verify the change works? What does the acceptance loop look like?

Read more
21

Artifacts of Success & Failure

What evidence gets produced? Screenshots, logs, test results, diffs — how do you capture and store proof?

Read more
22

PR Review in Practice

What does a real agent-authored PR review look like? Who reviews what, and when does a human step in?

Read more
23

Building with the Claude Agent SDK

How do you use the Claude Agent SDK to wire up agentic workflows? Patterns, gotchas, and practical examples.

Read more

Who This Is For

This site is for:

Roles

  • Staff and principal engineers
  • Platform / infra engineers
  • Engineering leads
  • CTOs and founders

Responsibilities

  • Development environments
  • CI / CD systems
  • Code review workflows
  • Secrets and identity
  • Governance and compliance

Bookmark this

You'll come back here.

Every team building an agentic dev system will hit the same questions — about environments, identity, review, governance, and trust. This is the reference that has the answers.