Skip to main content
1

Create a Project

​
A Project is the workspace boundary for a codebase.

  • Add your project folder. This defines the sandbox the agent can safely read from and write to.
  • (Optional) Add an Agent Primer to set default context for every new session.
    • Use it for code conventions, naming patterns, architecture notes, workflows, product goals, and “gotchas”.
    • Treat it like the one page you wish every new teammate read before touching the repo.
  • blprnt will build context over time.
    • The agent can create memories to preserve key decisions, terminology, and structure.
    • Future sessions can reference these memories to stay consistent without you repeating yourself.
    • Memories can include user preferences too (e.g., “avoid red, use blue instead”).
2

Start a Session

​
Sessions are where you collaborate with the agent to get real work done.

  • Start a new Session inside your Project.
  • Describe what you want in a few sentences, including your constraints (scope, quality bar, and any “don’t touch” areas).
  • If you have context that shouldn’t be forgotten, add it to the prompt or to the Agent Primer so it applies to future sessions too.
3

Answer Clarifying Questions (If Needed)

​
If the request has multiple valid interpretations, blprnt will pause and ask you targeted multiple-choice questions.

  • This avoids wasted work and wrong assumptions.
  • Your answers become the basis for the plan and the acceptance criteria.
  • If something feels missing, this is the moment to add constraints (performance, UX expectations, compatibility, timelines).
4

Review the Plan & Todos

​
Planning happens automatically: the agent will propose a plan with clear todos and acceptance criteria.

  • You’ll see a written plan that describes the approach and defines what “done” means.
  • You’ll see a list of todos with statuses so progress is visible as work happens.
  • Review for scope and sequencing:
    • If it’s too big, ask to split it into smaller todos.
    • If it’s too risky, ask for safer steps (or a research-first todo).
    • If it’s missing something, ask to add acceptance criteria before execution starts.
5

Watch Execution & Verification

​
Once you approve, the agent will move through the todos and keep the plan updated as it goes.

  • The primary agent may spawn specialized subagents to handle parts of the work (research, design, execution, verification).
  • Todos move through statuses as each unit of work is completed and checked.
  • If verification finds an issue, the agent will loop back to fix it before continuing-so you don’t accumulate broken work.

​
Suggested First Task (Safe + Useful)

If you’re not sure what to try first:
Read through the codebase and understand the overall architecture and flow. Then suggest a small UI improvement with a plan to implement it.
Or if it’s a brand new project, you can start with a brainstorming session to get a sense of the project and then create a plan to implement it. It’s the fastest way to learn the blprnt loop: plan → execute → verify.
If you’re not sure what to do, just ask the agent to help you.