# Write your team's process once. Every agent runs it.

Define your team's workflows once, with shared stage prompts and skills. The server enforces their gates, and you see where every task is.

Source: https://codeherder.com/team-process/

Give your coding agents a shared delivery process. Define workflows with the team's stage prompts and skills. CodeHerder enforces each task's required gates and shows you where the work is.

[Request access →](https://codeherder.com/get-started/#request-access) [See what the team defines →](https://codeherder.com/team-process/#define-once)

## Ask five engineers how they would add SSO

Five engineers can share a repo and still run five different agent processes.

### What that costs the team

- Shared instructions do not show whether each agent followed them.
- Checks and hand-offs depend on the person running the agent.
- You piece together progress across chats and terminals.
- You cannot compare approaches, because no two tasks ran the same way.

1. Engineer 1 One long chat with one agent. Reads the whole diff at the end.
2. Engineer 2 Three agents in three terminals. Merges whatever passes locally.
3. Engineer 3 A plan-first prompt kept in a personal notes file.
4. Engineer 4 Their own skills folder and a review checklist only they use.
5. Engineer 5 Writes the code by hand and asks an agent for the tests.

With CodeHerder: the team's story workflow, shared by every engineer

1. plan
2. code
3. review
4. merge
5. verify

Each stage uses the team's prompt and skills. The server checks its gates before the task moves on.

## Define the process once

The team writes down how work gets delivered. CodeHerder stores it in the workspace and runs every task through it.

- The workflow A workflow for each kind of work, such as a story or a bug. Each stage says what it must produce before the task moves on.
- Stage prompts Each stage's prompt lives in the workspace's stage library. Every workflow built from that stage uses the same prompt.
- Skills One catalog of versioned skills for the workspace. A stage can narrow which skills its own sessions get.
- Shared memory A workspace wiki that every new session can read. A lesson one task learns reaches the next one.
- Checks and gates Each workflow's required checks and approval gates run on the server. An agent cannot skip one, whoever filed the task.
- Guardrails A composition policy can require stages and checks in pipelines built from the library. It can also limit which model tiers they use.

Every plan starts with built-in workflows, shared the same way. Writing your own workflows and stage prompts needs Starter or higher. See [pricing](https://codeherder.com/pricing).

## It reaches every engineer

Engineers still file tasks their own way: from the web app, the `ch` CLI, or claude.ai and ChatGPT. What happens next is the same for all of them.

1. Set it in a group Put workflows and skills in a group workspace. Every team workspace under it inherits them.
2. Change it in one place Save a shared stage to update the workflows built from it in that workspace. The preview shows which workflows change and which team-owned copies need a refresh.
3. New sessions pick it up Each new session uses the workflow and enabled skills that apply to its task. Nobody has to copy a file.

## See where every task is

See each task's stage and owner, and what it has cost, in one place.

- Where each task is The task board shows every task by stage, with its owner. The stalled count shows any task that stopped moving.
- What it cost Each task shows recorded spend, stage by stage, rework included.
- What the agent is doing Open any session's live terminal. Answer its question, or take over.
- What happened Each stage leaves a hand-off note. The activity feed records each status change and each blocker.
- What it ran with Check that a skill reached a session, and which version of each stage a task used.

Read more about [activity feeds](https://codeherder.com/docs/activity/) and [how work flows](https://codeherder.com/docs/how-work-flows/).

## Improve the process, not each person's habits

One process gives you one thing to measure and one place to change it.

- Propose a change Any member, or an agent working a task, can propose a workflow change. An owner or admin applies it or rejects it.
- Keep the history Every stage save is a new version, with who made it and why. Revert to an earlier version from the CLI.
- Test before you switch Replay real finished tasks on the old and new setup. Compare how often each passes first time, and what each costs. Nothing switches on its own.

See [the stage library](https://codeherder.com/docs/stage-library/), [workflow proposals](https://codeherder.com/docs/workflow-proposals/) and [skills](https://codeherder.com/docs/skills/).

## Start with one team and one repo

Begin with the built-in story workflow. Turn on your required checks. Run two weeks of real tasks, then change one stage at a time and compare.

[Read the two-week rollout plan →](https://codeherder.com/rollout) [How CodeHerder works →](https://codeherder.com/how-it-works) [For leaders →](https://codeherder.com/for-leaders)
