Getting Started
Install, initialize, and run your first change end to end.
Installation
agentic-fy is a Node.js CLI. You need version 20.19.0 or newer. Check your Node:
node --version
Install the CLI globally:
npm install -g @agentic-fy/agentic-fy
If you can't install globally (permission error under /usr/local), run it directly without installing:
npx @agentic-fy/agentic-fy --version
Verify it worked
agentic-fy --version
If it prints the logo and a version number, the CLI is on your PATH.
Your first five minutes
The whole loop:
$ npm install -g @agentic-fy/agentic-fy
$ cd your-project && agentic-fy init
$ agentic-fy explore (optional: think it through first)
$ agentic-fy propose add-dark-mode (drafts the plan; you review it)
$ agentic-fy apply (tracks the tasks)
$ agentic-fy verify (checks it's ready)
$ agentic-fy archive (change archived)
agentic-fy explore. It's a no-stakes
thinking partner: it shows the project state and active changes, and helps turn a fuzzy idea into a
concrete plan before any code is written.
What agentic-fy creates
After running agentic-fy init, your project gets this structure:
agentic-fy.config.yaml # project configuration
agentic-fy/
├── specs/ # project specs
├── changes/ # proposed changes (one folder per change)
│ └── archive/ # completed changes
The init command creates the base structure idempotently (running it again does not
overwrite the existing agentic-fy.config.yaml) and configures the MCP integration for the
AI tools you select. Pass --tools kiro,cursor to skip the prompt, or --tools none
for just the base structure.
When you create a change with propose, it looks like this:
agentic-fy/changes/<change-name>/
├── proposal.md
├── design.md
├── tasks.md
├── specs/
│ └── <change-name>.delta.yaml
└── .agentic-fy.yaml # metadata (name, status, dates)
Understanding the artifacts
| Artifact | Purpose |
|---|---|
proposal.md | The "why" and "what" — intent, scope, and approach |
specs/<capability>.delta.yaml | Spec delta: how the requirements change (merged into the project specs on archive) |
design.md | The "how" — technical approach and architecture decisions |
tasks.md | Implementation checklist with checkboxes |
Example: your first change
Let's SideBar Menu mode to an application.
1. Initialize the project
cd your-project
agentic-fy init
2. Create the change
$ agentic-fy propose add-dark-mode
[propose] add-dark-mode
Created: proposal.md
Created: design.md
Created: tasks.md
Created: specs/add-dark-mode.delta.yaml
3. Fill in the artifacts
Edit the files in agentic-fy/changes/add-dark-mode/. Write the real tasks in
tasks.md and describe the requirement changes in the spec delta:
# Tasks — add-dark-mode
- [ ] 1. Create ThemeContext with light/dark state
- [ ] 2. Add a theme toggle to settings
- [ ] 3. Persist the preference in localStorage
4. Track the implementation
$ agentic-fy apply
[apply] add-dark-mode
Applying change "add-dark-mode".
Tasks: 0 completed, 3 pending.
[ ] 1. Create ThemeContext with light/dark state
[ ] 2. Add a theme toggle to settings
[ ] 3. Persist the preference in localStorage
Implement the tasks and mark them as [x] in tasks.md as you go.
5. Verify
If your requirements declare a verify command in the spec delta, verify
runs them and reports evidence per requirement. A change is only marked verified when the
evidence passes and there are no gaps (use --allow-gaps to accept requirements without a
command).
$ agentic-fy verify
[verify] add-dark-mode
All artifacts present (proposal, design, tasks).
All tasks marked as completed.
Evidence: 2/2 requirement(s) proven.
✓ dark-mode-toggle (npm test -- theme)
✓ persist-preference (npm test -- persist)
Status updated to "verified".
6. Archive
Archiving merges the change's spec delta into the project's consolidated specs, then moves the
change to history. Preview the merge first with agentic-fy archive --dry-run.
$ agentic-fy archive
[archive] add-dark-mode
Spec "dark-mode" created: +1 ~0 -0
Change "add-dark-mode" archived at agentic-fy/changes/archive/add-dark-mode.
Letting the AI drive (MCP)
The commands above run in the terminal. To let your AI assistant drive the workflow, connect the MCP server:
agentic-fy mcp
It exposes the explore, propose, apply, verify,
merge, archive, list, show, validate,
status, and context tools to any MCP-compatible agent. See the
Commands page for setup in editors like Kiro.