1. Inception

Before the first story, agilemarkdown asks you to write down what the work is actually for; that document is the inception.

1.1 Initialize a project

agilemarkdown projects are git repos. They have a .am/ directory for derived state and one folder of stories at the root. am create-backlog makes both.

The board for this moment is your file system. After am create-backlog, you have a stories/ directory next to a .am/ directory in the project root, both filled with plain markdown ready for git.

$ cd projects/meadow-letters
$ git init
Initialized empty Git repository
$ am create-backlog
created backlog: stories
created .am/config.yaml

create_backlog does the same: creates the directory, seeds the config, returns the path. Equivalent to the CLI verb. Useful when an agent is bootstrapping a fresh project on a fresh repo.

Full envelope

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "create_backlog",
    "arguments": { "name": "stories" }
  }
}

1.2 Write the inception

am inception seeds .am/inception.md with prompts for the six sections agilemarkdown asks every project to articulate:

The user, a sentence about who the work is for
The goal, one sentence the team agrees with
The reason, why now, why this team
Success, what has to be true for the work to have mattered
Constraints, hard limits on time, scope, dependencies
Out of scope, what you're consciously not doing

You edit the file in any editor (VS Code, vim, the lab's Terminal tab, same file). When the document feels honest, you commit it. Stories come after.

The inception lives at .am/inception.md. Open it in any markdown editor, VS Code's preview, vim, the lab's Terminal tab. Same file, same content; no agilemarkdown-specific UI is needed because markdown is the UI.

$ am inception
seeded .am/inception.md (edit it; ask the team)

$ am inception --show
# Inception

## The user
(describe the people you are building for)

## The goal
(one sentence the team agrees with)

## The reason
(why now, why us)

## Success
(what specifically must be true for this to have worked)

## Constraints
- (hard limits on time, scope, dependencies)

## Out of scope
- (what you're consciously not doing)

inception_doc reads or writes the file. No args returns the current content (or the seeded template). Passing text: overwrites the file with what the agent thinks the project's inception should say.

Full envelope

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "inception_doc",
    "arguments": {}
  }
}

1.3 Record team agreements

Team agreements are anything the team has decided matters: "bugs at the top of priority," "no merges Friday afternoon," "questions in #project-meadow within a day." The coach warns when an action would break one but never refuses; the team can override with a reason.

$ am team-agreements --add "Bugs go to the top of priority"
recorded: 1 agreement
$ am team-agreements --add "8-point hard cap. Above 3 is a red flag."
recorded: 2 agreements
$ am team-agreements --add "PM rejects with a reason"
recorded: 3 agreements

team_agreements with add: appends. With no args it returns the current list. Useful for an agent to surface in a prompt like "the team has decided: …" whenever the agent is about to do something that touches a soft rule.

Full envelope

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "team_agreements",
    "arguments": {
      "add": "Bugs go to the top of priority"
    }
  }
}

What just happened

A new git repo became an agilemarkdown project (one am create-backlog + a .am/ directory)
The inception lives in a markdown file in .am/inception.md, open in any editor, commit with the rest of the repo
Team agreements live in .am/team-agreements.md, one bullet per agreement, dated
All three artifacts are plain files in git. No SaaS, no database, no central server.

Next: Create the first story →