Welcome to the Easy Claude Code Beta
Thank you for testing. This guide walks you through setup, testing protocols, and the feedback we need.
What is Easy Claude Code?
Easy Claude Code (ECC) is a remote execution dashboardfor Claude Code. It lets you send coding tasks from your browser — phone, tablet, or desktop — to Claude Code running on GitHub Actions runners. Claude reads your repo, writes the code, commits to GitHub, and triggers auto-deploy. Your laptop doesn't need to be on.
ECC is not Claude Code itself. It is a dashboard that dispatches tasks to Claude Code running in sandboxed cloud environments.
Prerequisites
Before you start, you'll need:
- A GitHub account with at least one repository you're willing to test with
- A test repository (see below)
- A modern browser (Chrome, Safari, Firefox)
No Anthropic API key needed — the beta includes free Claude credits so you can test without any setup. Power users can optionally add their own key for higher limits.
Recommended: Create a Test Repository
We strongly recommend testing with a disposable repo first — not your production code.
1. Go to github.com/new
2. Create a repo called "ecc-test" (public or private)
3. Add a simple project — a Next.js starter, a Python script, anything
4. Push at least one commit so the repo isn't empty
Setup (5 minutes)
Sign up
Go to the onboarding page, walk through the 4-step wizard, and create your account.
Add a project
From the dashboard, navigate to a project or create one. Enter your GitHub repo in owner/repo format. Optionally add your live URL.
Enable cloud execution
On the project detail page, click "Enable Cloud Execution". ECC automatically installs a GitHub Actions workflow and configures secrets. Beta testers get free Claude credits — no API key needed.
Send your first task
Type a task in the text area — plain English — and click Run. Watch the status update from Queued to Running to Completed.
Permission Tiers
ECC has three permission tiers. You choose a default during onboarding and can change it per-project anytime.
Locked
(Approve Everything)Every tool call needs your approval from the dashboard
Best for: Production repos, high-sensitivity code
Standard
(Smart Defaults)Safe ops (read, git, npm) auto-approved. Destructive actions need your OK
Best for: Most testing — recommended for beta
Autonomous
(Bypass Permissions)Full autonomy. Only hard safety limits apply
Best for: Disposable test repos, speed testing
For beta testing, we recommend starting with Standard, then trying Locked and Autonomous to compare the experience.
All tiers enforce a hard deny list that blocks:
Testing Protocols
Five structured rounds. Work through them in order — each builds on the previous.
Onboarding Flow
~10 minutesWalk through the full signup flow as if you've never seen the product.
- Does the 4-step wizard make sense? Is each step clear?
- Do the Back/Next buttons work correctly?
- Is the permission tier selector understandable without explanation?
- Does the Autonomous tier warning feel appropriately serious?
- Does the magic link email arrive promptly?
- After signing in, do you land on the dashboard?
Project Setup
~10 minutesConnect a GitHub repo and enable cloud execution.
- Can you add a project and connect a GitHub repo?
- Does the platform selector (GitHub/GitLab/etc.) work?
- Does the one-click setup (workflow install + secret provisioning) complete?
- Is the progress feedback clear (checkmarks, error messages)?
- Does the "Cloud execution ready" banner appear?
Task Execution
~15 minutesSend real tasks and evaluate the results.
Try these task types:
- "Create a file called hello.txt with the text 'Hello from ECC'"
- "Add a console.log to the main entry point"
- "Add a footer component and import it in the layout"
- "Fix any lint or type errors in the project"
- "Add a dark mode toggle using CSS custom properties"
- Does the task get picked up by GitHub Actions?
- Is the status progression (Queued > Running > Completed) visible?
- Does the diff viewer show what changed?
- Is the commit visible on GitHub?
- If you have auto-deploy, does the "View Live" button work?
- For failed tasks, is the error message useful?
Permission Tiers
~15 minutesTest each tier to understand the differences.
Locked mode:
- Switch a project to Locked and send a task
- Do permission prompts appear in the dashboard?
- Can you approve/deny individual tool calls?
- Does "Always" remember the decision for future calls?
Standard mode:
- Do safe operations run without prompts?
- Try a task involving file deletion — does it ask for approval?
Autonomous mode:
- Does the confirmation warning appear when selecting?
- Send a task — does it run end-to-end without prompts?
Dashboard & Navigation
~10 minutesEvaluate the overall experience.
- Does the Morning Briefing show your projects with correct status?
- Do permission tier badges appear on project cards?
- Do the "next step" links guide you correctly?
- Does the landing page accurately describe what ECC does?
- Is the mobile experience usable? (Try from your phone)
Feedback We Need
High priority
Clarity
Did you understand what ECC does and how it works without external explanation? Where did you get confused?
Permission model
Does the 3-tier system make sense? Did you feel in control? Was the Autonomous warning clear enough?
Trust
At any point, did you feel uncomfortable about what ECC might do to your repo? What would make you feel safer?
Friction points
Where did the flow feel slow, confusing, or broken? What made you hesitate?
Task quality
How good were the code changes Claude made? Were they committed correctly?
Medium priority
Mobile experience
How was the phone experience? Could you realistically use this on the go?
Error handling
When things went wrong, were error messages helpful? Did you know what to do next?
Onboarding
Was the 4-step wizard the right amount of information? Too much? Too little?
Nice to have
Feature requests
What's missing? What would make you use this daily?
Naming and copy
Does "Easy Claude Code" make sense as a name? Is the language on each page clear?
Comparison
How does this compare to using Claude Code directly in the terminal? When would you choose ECC over the CLI?
How to Submit Feedback
Please include:
- Which testing round you were on
- What happened (screenshots are very helpful)
- What you expected to happen
- Device/browser you were using
- Your permission tier at the time
Known Limitations
Troubleshooting
| Problem | Fix |
|---|---|
| Task stuck on "Queued" | Check GitHub Actions tab on your repo — enable workflows if prompted |
| "Workflow not found" error | Go to your repo's Actions tab and enable workflows |
| API key rejected | Verify the key starts with sk-ant- and has not expired |
| No diff shown | Claude may not have made changes — check the "no changes" indicator |
| Magic link not arriving | Check spam folder. Try a different email if persistent |
A Note from Dennis the Founder
Thank you for testing. Your feedback directly shapes the product.