What stops an agent from obeying text it reads?
An agent reads text from many places, and some of it can carry an instruction. What CodeHerder limits and what stays open. Then the settings to turn on.
A task is not done when an agent stops typing. It is done when it passes review, merges, and ships.
A workflow is the sequence of stages a task moves through, from a backlog item to a merged change. CodeHerder ships with a starter workflow and lets you edit every stage: who reviews, what gates a merge, and which stages need a human at all.
Most teams' first workflow is too loose. An agent commits straight to a shared branch, or a review stage runs with no rubric. Results then depend on which agent happened to pick up the task. A workflow you can inspect and edit fixes that, because the rules live in the system, not in one person's head.
This hub is for anyone setting up a workflow, or tuning one they already run. It covers what a stage should check, and when to add a human gate. It also covers what teams changed after their first week of real use.
Request access →An agent reads text from many places, and some of it can carry an instruction. What CodeHerder limits and what stays open. Then the settings to turn on.
A big ticket fails for agents and for people. How epics, child tasks and dependencies turn one idea into small changes that merge in order.
One report says keep AI code to 25-40%. Our own repo is at 97.5%. Review capacity sets the limit, not code share. Here is how to find yours.
Each task gets a Slack card and a thread. Approve, answer or unblock with a button, or reply in the thread.
A plugin adds actions to your account. Agents call them as tools. Under the default policy, no write runs until a person approves it.
A change can be correct and still wrong for the task. Here is what our review agent reads, what it must do, and where it stops.
An agent wrote the change. Who signs it off? Here is how an approval gate keeps the author and the approver apart in CodeHerder.
Each acceptance item names the stage that proves it. An unticked item blocks any move past that stage. Here is how proof stages work.
Agents write and review code faster than CI can check it. Here is the budget, the split and the loop that took CodeHerder to its best merge days ever.
One short, vague brief, run 84 times by agents. Median cost about $18, median time 90 minutes. Here is what shipped and what it cost.
Our first Agent Experience survey put five questions to every code-stage agent for a day. 84 of 85 answered. Here is how it works and what the herd told us.
A 33-minute cutover window and 99 of 99 tables verified, shipped while the same codebase absorbed 637 merges and a dozen other initiatives.
Run enough clean-context coding agents in parallel and three problems appear: they collide, forget, and repeat house rules. Here's how CodeHerder fixes that.
Keeping CLAUDE.md small is best practice, but ours hit 231KB across 37 correct commits. Splitting it cut cost per turn 28%, turns per task by a third.
CodeHerder now speaks remote MCP: check tasks, triage the backlog, file a task, and comment, straight from claude.ai or chatgpt.com.
944 user stories shipped in a week, with no human writing the code. Here's what the median one cost, and what each quality gate caught.
How to ship real software with AI agents when you do not write code: write a checklist brief, choose your approval gates, review the result and set a spend cap.
Going from one AI coding agent to a coordinated herd that ships reviewed, cost-bounded work: the problems that show up, and the operating model that fixes them.
A single coding agent is easy to supervise. Ten of them is a coordination problem nobody's tool was built to solve.
Follow one real plain-English bug report from filed to merged: the five stages, the questions agents ask, what it cost, and where you can add approval.
File a task. An agent builds it. A person reads every request. You see the cost of each task.