---
source: CrewKit documentation
url: https://crewkit.io/docs/blueprints.md
title: "Blueprints"
description: "A living project plan — epics and tasks linked to the real conversations that did the work."
---


A Blueprint is your project's living plan: epics broken into atomic tasks, with dependencies, acceptance criteria, and links to the actual sessions that did the work. It stays useful because it's connected to reality — not a doc that drifts out of date.

---

## Create a blueprint

### Dashboard

Go to **Projects** → your project → **Blueprints** and generate one from a goal. crewkit drafts epics and tasks with AI; review and edit before activating.

### CLI

```bash
crewkit blueprint create "Migrate billing to usage-based pricing"
crewkit blueprint create --from goal.md          # Read the goal from a file
crewkit blueprint create -a <artifact-id>        # Include an artifact as context
crewkit blueprint create --dry-run               # Generate and display without uploading
```

Generation runs with full codebase access, so the plan reflects the code you actually have.

---

## Epics, tasks, and dependencies

- **Epics** group related work into major workstreams.
- **Tasks** are atomic units with a status: `pending`, `in_progress`, `completed`, `blocked`, or `skipped`.
- **Dependencies** between tasks are tracked and visualized as a graph in the dashboard.
- **Acceptance criteria** live on each task as a checklist.

---

## Co-pilot chat

Each blueprint has an AI co-pilot in the dashboard. Chat with it to refine the plan — split an epic, tighten a task, question the ordering — and apply its suggestions directly.

---

## Work the plan from the CLI

```bash
crewkit blueprint list                       # Blueprints for the current project
crewkit blueprint show <id>                  # Detail with epics and tasks
crewkit blueprint tasks --status pending     # List tasks, optionally by status
crewkit blueprint start <task-id>            # Claim a task: sets in_progress + injects its context
crewkit blueprint complete <task-id>         # Mark a task completed
crewkit blueprint review                     # AI review of the plan against the local codebase
crewkit blueprint export                     # Export the plan to .claude/work/
```

`blueprint start` injects the task's context into your next session, so the agent knows exactly what it's working toward.

---

## Sessions link to tasks

When a project has an active blueprint, `crewkit code` detects it automatically and connects your session to the plan. Tasks accumulate links to the real conversations that did the work — open any task in the dashboard to see them.

Teammates see claimed tasks in the project's active-work view, so nobody starts the same task twice.

---

## In the terminal

In crewkit's interactive terminal interface (`crewkit tui`), the sidebar shows the active blueprint during a session: name, progress, task counts, and the task you're currently working on.

---

## Next steps

- [Artifacts and project memory](/docs/artifacts)
- [MCP server](/docs/mcp)
- [Session sharing](/docs/sharing)