Projects
Start your first project
Each project guides you through the full five-stage methodology, from problem definition to a working, documented solution.
New project
Name your project and describe what you are trying to build or improve. You can refine this as you go.
Different AI tools have different requirements. Knowing this upfront helps Hora guide you through the right decisions in later stages.
Coding agents (Claude Code, Cursor, Windsurf, Gemini Code Assist, GitHub Copilot, OpenAI Codex) work directly in your codebase through a terminal or an IDE extension like VS Code. It's a more technical approach than chat AI or vibe coding platforms, but you don't need to figure it out alone: an AI chatbot can walk you through the setup step by step.
Delete project?
This will permanently delete the project and all its stage data, decisions, and notes. There is no undo.
Loading...
Before Stage 2, it helps to know how you plan to build this. You can update the build tool setting when creating a new project. Editing this on existing projects is coming in a future update.
of steps complete
Step of
This prompt distills your answers into a structured request. Copy it and paste it into Claude, ChatGPT, or any AI you use. The AI will return a Problem Definition Document you save in Step 4.
View prompt text
The best documents come from back-and-forth, not a single response. Push back on the AI, iterate on specific sections, then use the finalization prompt to get a clean final version.
How to get a better result
- Read the full response first, then pick one or two things that do not match your actual situation. Starting everywhere means improving nothing.
- Be specific when you push back. "Section 2 feels generic" is not enough. Say: "Section 2 describes [X], but my situation is actually [Y]." The AI needs the delta, not just a complaint.
- Ask it to challenge you: "Is there anything in my inputs that seems inconsistent, or that I should reconsider?" You want a collaborator, not a transcription service.
- Work on sections, not the whole document. Paste the section you want to revise, describe the problem, and ask for a rewrite of that section only.
- If the AI keeps asking clarifying questions instead of producing output, say: "Make a reasonable assumption and keep going." If it drifts into tangents, say: "Stop — focus on the document only."
- When you are satisfied, copy the finalization prompt below to get a clean standalone version you can save.
How much did you workshop this?
Reflection note (optional)
Paste your AI response here, or upload the file if you saved it as .txt or .md.
If you talked through your problem definition with an AI, paste this prompt at the end of that conversation to get a structured decision summary. Then paste the response below to save it to your Decision Journal.
For best results, paste the AI's numbered list directly. If typing manually, use a numbered or bulleted list with one decision per line.
Paste your Problem Definition Document above to finish Stage 1.
Loading...
of steps complete
Step of
This prompt combines your Stage 1 problem definition with your Stage 2 inputs. Copy it and paste it into Claude, ChatGPT, or any AI you use. The AI will return a Constraint and Scope Document you save in Step 4.
View prompt text
The best documents come from back-and-forth, not a single response. Push back on the AI, iterate on specific sections, then use the finalization prompt to get a clean final version.
How to get a better result
- Read the full response first, then pick one or two things that do not match your actual situation. Starting everywhere means improving nothing.
- Be specific when you push back. "Section 2 feels generic" is not enough. Say: "Section 2 describes [X], but my situation is actually [Y]." The AI needs the delta, not just a complaint.
- Ask it to challenge you: "Is there anything in my inputs that seems inconsistent, or that I should reconsider?" You want a collaborator, not a transcription service.
- Work on sections, not the whole document. Paste the section you want to revise, describe the problem, and ask for a rewrite of that section only.
- If the AI keeps asking clarifying questions instead of producing output, say: "Make a reasonable assumption and keep going." If it drifts into tangents, say: "Stop — focus on the document only."
- When you are satisfied, copy the finalization prompt below to get a clean standalone version you can save.
How much did you workshop this?
Reflection note (optional)
Paste your AI response here, or upload the file if you saved it as .txt or .md.
If you worked through your constraints with an AI, paste this prompt at the end of that conversation to get a structured decision summary. Then paste the response below to save it to your Decision Journal.
For best results, paste the AI's numbered list directly. If typing manually, use a numbered or bulleted list with one decision per line.
Paste your Constraint and Scope Document above to finish Stage 2.
Loading...
Before we start, let's confirm how you plan to build this.
Stage 3 questions are tailored to your build approach. You marked this as undecided during project setup. Select your approach below to continue.
Not sure which one fits? Use the section below.
You can also change this any time in .
Answer these three questions. Copy the prompt, paste it into any AI tool, and come back with a recommendation.
Not sure what these tools are? Read the Build Tool Guide.
Paste this into Claude, ChatGPT, or Gemini. Bring the recommendation back and select your tool above.
of steps complete
Step of
This prompt includes your full Stage 1 and Stage 2 documents along with your Stage 3 inputs. Architecture conversations take multiple rounds. Copy the prompt, paste it into your AI tool, work through it together, then use the finalization prompt in Step 3 when you are ready to produce the final document.
View prompt text
Architecture decisions are the highest-leverage decisions in a build. Do not accept the first response as final. Push back, ask why, challenge assumptions, then use the finalization prompt to get a clean document.
How to get a better result
- Ask why, not just what. When the AI recommends a specific technology or pattern, ask why it fits your project better than the alternatives. A recommendation it cannot justify specifically is not grounded in your situation.
- Use your Stage 2 constraints as a filter. Before accepting any recommendation, check it against your hard constraints and skill level. If something conflicts, say so directly and ask for an alternative.
- Challenge the first answer. Ask what it would recommend differently if your skill level were higher or lower, or if your deployment requirement changed. The delta tells you what assumptions the recommendation is built on.
- Ask what you have not considered. "Is there anything in my inputs that seems inconsistent, or a risk I have not named?" You want a technical advisor, not a transcript of your own assumptions.
- Before finalizing, ask: if this decision turns out to be wrong, what would need to be rebuilt? The answer tells you which decisions carry the most risk. Focus your iteration there.
- If the AI goes deep on one technology and loses the thread of your actual question, say: "Stop. My question was [X]. Answer that first, then we can explore the rest." Architecture conversations can sprawl — it is fine to redirect.
How much did you workshop this?
Reflection note (optional)
Paste your finalized AI response here, or upload the file if you saved it as .txt or .md.
If you worked through your architecture with an AI, paste this prompt at the end of that conversation to get a structured decision summary. Then paste the response below to save it to your Decision Journal.
For best results, paste the AI's numbered list directly. If typing manually, use a numbered or bulleted list with one decision per line.
Paste your Architecture and Tech Stack Document above to finish Stage 3.
Loading...
These are your key answers from Stages 1 and 2. Review them — if anything has changed or you want to refine something before generating your Build Plan, edit it now. Then confirm to continue.
This prompt assembles your full project context and asks your AI tool to produce a phased Build Plan with feature lists, acceptance criteria, and risk callouts. Copy it and paste it into Claude, ChatGPT, or any AI you use.
View prompt text
Push back on the AI, challenge the phase structure, then use the finalization prompt to get a clean final version.
How to get a better Build Plan
- Ask why the phases are ordered this way. If the reason is not convincing, push for a different sequence.
- Check each phase against your Stage 2 constraints. If a phase implies something your constraints rule out, flag it explicitly.
- Challenge phases that seem too large. A phase with five features is probably two phases. Smaller phases mean faster feedback loops.
- Ask: "What is the riskiest assumption in this plan, and what happens if it is wrong?" A good AI collaborator will surface this; a mediocre one will not.
- If the AI suggests expanding scope mid-conversation, say: "Log that as a future consideration. Let's keep this plan to what I described." Some AI tools will keep adding features if you let them.
- When you are satisfied with the structure, use the finalization prompt to get a clean standalone version.
How much did you workshop this?
Reflection note (optional)
Paste your AI response here, or upload the file if you saved it as .txt or .md.
At the end of your AI conversation, paste this prompt to get a structured summary of every decision you made. Then paste the AI's response below to save them to your Decision Journal.
For best results, paste the AI's numbered list directly. If typing manually, use a numbered or bulleted list with one decision per line.
Loading...
Before your first session, let's confirm your build tool and get your AI collaborator oriented on your project.
A few quick questions so the setup prompt can give you the right guidance.
This is separate from the tool you use to build. Your finished product may itself call AI APIs — for example, if it summarizes text, generates content, or uses AI to process data.
This prompt gives an AI chatbot your full project context and asks it to walk you through setup. Take it to any AI chat tool, follow the guidance, then come back here to continue.
Preview prompt
The prompt asks the AI to provide a numbered decision list at the end. When you return, paste that list below to capture your setup decisions in your project journal.
Add this file to your project so your coding agent always has full context on what you're building, why, and where things stand. You won't need to re-explain anything session to session.
Your Project Context Document is saved in the Documents section of this app and updates automatically after each session. It's pre-loaded into your session briefings, so you don't need to paste it manually.
Something went wrong generating the document. Check the browser console for details, then try again.
Generating your context document…
Preview document
Hora parsed your Stage 4 Build Plan into the phases and items below. This is a read-only preview — confirm to activate your interactive checklist, then add, rename, or remove items freely in Plan View.
You can edit, add, or remove phases and features after confirming.
No Build Plan was found from Stage 4. Complete Stage 4 first, or start with an empty plan and add phases manually.
Ready to build? Go to the to start your first session. Come back here to mark features complete and keep your plan up to date as you go.
Project Context Document
-
Mark as changed
Remove this feature
Set up your plan first. Go to the Plan tab to confirm your Build Plan.
Copy this prompt to start your AI session. The pre-session check at the end asks your AI to flag anything worth addressing before you begin.
The briefing ends with a pre-session question. Read the AI's response before diving in — it often surfaces a risk or gap worth addressing first.
View briefing text
Keeping the AI on track
Some AI tools (ChatGPT in particular) are tuned for engagement. They may ask increasingly granular questions, suggest scope you did not ask for, or explore tangents that feel productive but pull you away from your session goal.
A few things that help:
- If it asks a clarifying question you cannot answer, say: "Make a reasonable assumption and proceed."
- If it goes down a tangent, say: "Stop. Let's return to the session goal: [your goal]."
- If it keeps asking follow-ups instead of building, say: "I need output, not more questions. Build it now and we can adjust."
- If it suggests adding scope mid-session, say: "Log that as a future consideration and continue with what we planned."
Your session briefing already frames the goal. You are allowed to hold the AI to it.
Copy this prompt, paste it into your AI session, then paste the response back here. Hora will parse the decisions and log them to your Decision Journal.
Select the phase you are working on, choose any specific items for this session, then describe your goal. Hora generates a session briefing that includes your full project context document — paste it into your AI tool to start a fully briefed session without navigating away.
Select specific items or leave blank to include all incomplete items from this phase in your briefing. Completed items are not shown.
What outcome are you driving toward — what will be true at the end of this session that isn't true now? This gives the AI the intent behind the checklist.
If the AI asks too many questions or heads off on a tangent, say: "Make a reasonable assumption and proceed" — or "Let's return to the session goal." Your session briefing frames the scope; you are allowed to hold the AI to it.
Project history
Loading...
Nothing logged yet. Complete a session, log a decision, or update a plan feature to start your project history.
Documents
AI-generated stage outputs and your reference files. Copy or download any document to use outside Hora.
Loading...
Preview
Reference Documents
Upload markdown or text files to keep reference material alongside your stage outputs — specs, research notes, constraints docs from outside Hora.
No reference documents yet. Upload a .md or .txt file to get started.
Preview
Decision Journal
Every decision recorded across all stages — from AI conversations, session ends, and manual entries. Newest first.
Loading...
No decisions yet
Decisions are logged automatically at the end of each stage and session. Use the Log Decision button in the sidebar to add one manually.
No decisions match the current filter.
Notes
Free-form notes attached to this project. Capture anything worth keeping outside of decisions and sessions.
Loading...
No notes yet
Add free-form notes to capture anything that doesn't fit a decision or session log.
No notes match your search.
Build Tool Guide
Each approach works differently. The right choice depends on what you are building, where it will run, and how much control you need.
Chat AI
Claude.ai, ChatGPT, GeminiNo setup, no code. You are not writing software. You are designing a system that runs in conversation with an AI tool. This might be a prompt library, a structured analytical framework, a repeatable decision-making workflow, or a document system. The product is a method, not an application.
Works with any AI subscription, including free tiers. No local setup required.
Examples
- A competitive analysis framework you run once a month
- A structured prompt system for reviewing content before publishing
- A weekly planning workflow with built-in reflection questions
Vibe coding platform
Bolt.new, Lovable, ReplitBrowser-based, no local setup. You build by describing what you want. The platform generates the code and runs it in the browser. No terminal, no file system. Most useful for form-based apps, dashboards, and simple data tools where the interface matters.
Works best when the data model is relatively simple and the app does not need deep control over back-end logic. The ceiling is lower than a custom-coded solution, but you can ship something working in hours.
Examples
- A client-facing intake form that routes to a review dashboard
- A simple inventory tracker with a clean interface
- A shared team task board with status tracking
Coding agent
Claude Code, Cursor, Windsurf, Gemini Code AssistWorks in a local code editor. The agent reads and edits your actual files. This gives you complete control over the tech stack, data model, and build process. It is more technical than vibe coding, but an AI chatbot can walk you through the initial setup. No separate developer API subscription is required for most entry points.
The tradeoff: you need a local development environment (a terminal, installed software, and access to the file system). If you have never used one, there is a short learning curve upfront.
Examples
- A financial planning tool with a local database and Monte Carlo simulation
- A custom data pipeline that pulls from APIs and runs locally
- A Python web application with a full back end and structured data model
No-code tools
Airtable, Notion, Zapier, Make, GlideNo code at all. These platforms give you a visual interface for building databases, automations, and simple apps. The ceiling is lower than a custom-coded solution, but setup is faster and maintenance is simpler. The right choice when your problem fits what these platforms already do well.
The main risk: discovering mid-build that your requirements exceed what the platform can handle. Be honest about that ceiling before you start.
Examples
- A client project tracker in Airtable with automated status notifications
- A Notion workspace with a structured intake form and review workflow
- A Zapier automation that routes new leads to the right team member based on criteria
Hora's stage outputs work with all of these approaches. For coding agent projects, Stages 3 through 5 produce briefing documents you bring into Claude Code or Cursor. For all other approaches, they give you the context you need for your next AI conversation.
Project Settings
Update project details and build configuration.
Hora uses this to tailor questions and guidance in Stage 3 and beyond. Changing it here re-routes future stages but does not alter work already done.
Coding agents (Claude Code, Cursor, Windsurf, Gemini Code Assist, GitHub Copilot, OpenAI Codex) work directly in your codebase through a terminal or an IDE extension like VS Code. It is a more technical approach than chat AI or vibe coding platforms, but you do not need to figure it out alone: an AI chatbot can walk you through the setup step by step.
Danger zone
Delete project?
This will permanently delete the project and all its stage data, decisions, and notes. There is no undo.
Help
Everything you need to use Hora effectively.
Settings
Manage your data and preferences.
Data backup
All data is stored locally on this device. Export a backup regularly to protect against browser storage being cleared.
Last backup: No backup on record.
AI integration
By default, Hora generates a context-aware prompt that you open in Claude or ChatGPT. If you have an Anthropic or OpenAI API key, you can enable in-app AI without switching tabs.
Coming soonInstall Hora
Add Hora to your desktop or home screen for quick access without opening a browser tab.
Installing Hora is closer to saving a bookmark than downloading software. The browser adds a shortcut to your desktop or home screen. No files are installed on your computer. Your data stays exactly where it is now: on your device, in your browser. If you change your mind, remove it the same way you'd remove any app shortcut.