Reverse-Engineer Startup Growth

Use an AI agent to create a data-backed growth blueprint for a startup.

Introduction

30 Second Summary

A startup can seem to grow overnight. The real reasons are often scattered across the web.

In this project, you will use Claude Code with the treg plugin to reverse-engineer how a company grew. The rea-growth skill turns paid data calls into a Markdown blueprint with growth plays you can evaluate.

What You'll Build

The finished run gives you a growth blueprint that turns scattered evidence into a company story you can inspect.

By the end of this project, you'll have:

A growth blueprint showing the company's estimated traffic sources. A separate section captures its advertising signals.
An evidence trail built from backlinks. Social posts show where online attention gathered.
A copyable playbook with ten growth moves. Each move includes a prompt. A cost breakdown shows what the supporting calls cost. Your own note identifies the plays you would try.
An optional Secret Mission to push your growth research skills further.

Are there any prerequisites?

You need access to a terminal. Claude Code or another AI agent that supports skills connects to your treg account.

Before We Start

Before the hands-on work begins, lock in the growth blueprint you are about to create and why its evidence matters.

Install and Log In to treg

Your future growth teardown depends on data that treg can reach from the terminal. The CLI must be available before any catalog search can run.

You will authenticate the account that covers metered calls. A balance check will confirm the one-time $1.00 starter credit before you spend any of it.

In this step, get ready to:

Install the treg CLI from the official README.
Authenticate your treg account from the terminal.
Confirm the one-time $1.00 starter balance.

Install the treg CLI

The official treg README provides a shell installer for the CLI. The installer points the command at the hosted registry.

Running a remote installer is a security-sensitive action. The official README is the source of truth for this command.

Open the official treg README to locate the Quickstart installer.
Install the treg CLI by running this command:

curl -fsSL https://treg.to/install.sh | sh

What does this command do?

The curl command downloads the installer over HTTPS.
The -fsSL flags keep the download focused on a successful response.
The pipe sends the downloaded installer to sh.
The installer makes treg available in your shell.

The installer prints a completion message when the command becomes available. It also points you to the login command.

Check that your terminal can find the CLI by running:

treg --help

What does this command do?

The help check asks the installed CLI to list its commands. It does not change your account or spend any credit.

You should see help output for the treg CLI. The output includes commands for login, catalog access, calls, and balance checks.

✔️ I see the treg help output

Good progress. The treg command is installed and available from this terminal.

ⓧ The treg command is not found

A command-not-found result means your shell cannot locate the executable. The README also documents a plain Python package installation.

Install the plain CLI package by running:

pip install tools-registry

What does this command do?

This installs the tools-registry package. The package provides the treg command for the hosted registry.

Repeat the availability check by running:

treg --help

What should I see?

The command should now print the treg help output. That confirms the executable is available from your terminal.

Still can't find the command?

Your Python installation may place command-line tools outside the terminal's search path. Restart your terminal session after the package installation.

If the command remains unavailable, help me troubleshoot the treg installation.

Log in to your treg account

Authentication connects the CLI to your existing treg account. The default login flow uses GitHub in your browser.

Start the browser-based sign-in flow by running:

treg login

What does this command do?

The command starts treg's browser sign-in flow.
A successful sign-in stores an identity token locally.
Future treg commands use that identity to access your active account.

Complete the treg sign-in flow in the browser that opens.
Return to your terminal after the browser confirms the sign-in.
Decline guided onboarding if the CLI offers it.

The terminal confirms that the CLI is authenticated with your treg account. Your next commands can now access account-level information.

Login not completing?

Confirm that your terminal has an internet connection. Check whether the browser sign-in is waiting in another tab or window.

If the flow still does not finish, help me troubleshoot the treg login flow.

Confirm your starter balance

The balance view shows the prepaid credit available for metered calls. It also confirms that the CLI can read your authenticated account.

Before you check, do you expect the full starter balance to be untouched?

Check your authenticated account and prepaid credit by running:

treg balance

What does this command do?

The command reads the credit left in your active treg account.
It also reports recent spending from metered calls.
Checking the balance does not make a paid data call.

You should see your account with the one-time $1.00 starter balance. That confirms the CLI is authenticated before your first catalog call.

Your starting point is secured. The treg CLI can now reach the account that funds your growth research.

Balance missing or different?

Confirm that the terminal shows the treg account you intended to use. A balance tied to another active account may show a different amount.

If the account details remain unclear, help me diagnose my treg balance output.

Your CLI is installed and your starter credit is visible. Next, you will search the treg catalog for backlink tools and compare their per-call prices.

Find a Backlink Data Tool

Your treg account is funded with its one-time starter balance. The next question is which data source can reveal who links to the company you study.

Those incoming links are backlinks. The treg catalog shows the available data tools with a price for each call.

In this step, get ready to:

Search the treg catalog for backlink-data tools.
Review the available backlink-data tools.
Compare their per-call prices.

Search the backlink catalog

A per-call price is the amount deducted when a tool request runs. Checking that price first helps you plan how to use your starter balance.

Before you run the search, do you expect every backlink tool to have the same per-call price?

Search the catalog for backlink-data tools by running this command:

treg catalog search backlinks

What does this command do?

The catalog search asks treg to look through its registered data tools.
The backlinks term narrows the results to tools that provide backlink data.
The returned entries include per-call prices before you make a data request.

You'll see a list of available backlink-data tools. Each result shows its per-call price.

Catalog Search Not Returning Results?

Help me troubleshoot this catalog search. Review the treg repository for current CLI guidance.

Return to the authenticated terminal session from the previous step.
Check your network connection if the catalog does not respond.

Compare the listed prices

The search results give you a priced menu of backlink-data tools. Reviewing the prices helps you understand the cost of each possible request.

Review the name of each backlink-data tool in the terminal results.
Locate the per-call price shown for each tool.
Compare the prices across the results.

Good work. Your terminal now pairs the available backlink-data tools with their per-call prices.

Your priced catalog now shows which backlink data tools are available. Next, you'll use one of those tools in a direct data request.

Make a Direct Data Call

Your catalog search showed several ways to inspect a company's backlink profile. The next proof is turning one priced endpoint into real data.

A direct treg call lets you see the provider's raw JSON before an agent interprets it. Checking the balance afterward proves how the prepaid charge is recorded.

In this step, get ready to:

Inspect one backlink tool's parameters and fixed per-call price.
Run a direct data call for a chosen company domain.
Confirm that the prepaid balance decreased by the call's price.

Inspect the selected tool

Each catalog endpoint has its own parameters. Its detail record shows the fixed price before you spend any balance.

Return to the backlink catalog results from earlier.
Choose one backlink endpoint with a fixed per-call price below your current balance.
Record its endpoint ID here: selected backlink tool endpoint ID.
Inspect the selected endpoint by running this command:

treg catalog get [[BACKLINK_TOOL_ID="selected backlink tool endpoint ID"]]

What does this command do?

The detail lookup retrieves the selected endpoint's documentation.
The output lists the parameters that the provider expects.
The output confirms the price before a paid data request is made.
The ready call line shows the argument format for this endpoint.

You'll see the endpoint documentation with its parameters. You'll also see its price and a ready call line.

Record the listed charge amount here: selected tool charge amount.

Can't see the tool details?

Check that the endpoint ID matches the catalog result exactly. A missing segment or extra space can point to a tool that does not exist.

Confirm that your treg session from earlier is still active. Help me check this endpoint lookup.

Call the backlink endpoint

This next command spends from your prepaid balance. You already confirmed a fixed per-call price, so the amount is known before you run it.

Choose a company domain that you recognize.
Find the ready call line in the endpoint details.
Copy the arguments that follow the endpoint ID.
Replace the example target with your chosen company domain.
Record the completed argument string here: ready call arguments using your chosen domain.

Before you run the call, what kind of backlink fields do you expect this provider to return?

Make the paid data request by running this command:

treg call [[BACKLINK_TOOL_ID="selected backlink tool endpoint ID"]] [[CALL_ARGUMENTS="ready call arguments using your chosen domain"]]

What does this command do?

The call sends a request to your selected backlink endpoint.
The endpoint ID determines which provider handles the request.
The copied arguments preserve that provider's required input shape.
The response keeps the provider's data in its original JSON structure.

You'll see a raw JSON response for your chosen domain. The fields depend on the endpoint you selected.

Didn't get backlink data?

Compare the endpoint ID with the detail record. Confirm that every required argument from the ready call line is present.

Check that your company domain replaced only the example target. Help me diagnose this direct call.

Verify the balance change

The prepaid balance creates a visible record of the request cost. Comparing the new balance with the original starter balance confirms that the call was metered.

Before you check, what balance do you expect after subtracting selected tool charge amount from $1.00?

Check the remaining prepaid balance by running this command:

treg balance

What does this command show?

The balance check identifies the authenticated account.
The remaining amount reflects completed metered calls.
The difference from the starter balance gives you a concrete cost check.

You'll see a balance lower than $1.00 by selected tool charge amount. Your direct call now has both a data result and a verified cost.

Balance did not decrease as expected?

Confirm that the direct call returned data successfully. A failed request may not produce the expected charge.

Check that you selected an endpoint with a fixed platform per-call price. Help me compare the call price with my balance.

That's your first metered data retrieval complete. You now have raw backlink data plus proof of exactly what the request cost.

Your direct call proves that the catalog can return paid data on demand. Next, you will connect treg to Claude Code so an agent can use the same catalog during a growth teardown.

Connect treg to Claude Code

The raw JSON response proved that treg can fetch paid backlink data for you. Your updated balance showed the exact cost of that call.

That success still lives in the terminal. A Claude Code plugin brings treg into your agent. The rea-growth skill then gives the agent a repeatable process for growth research.

In this step, get ready to:

Install the treg marketplace plugin in Claude Code.
Connect the plugin to your treg account.
Make rea-growth available in the agent's skill list.

Install the treg plugin

Claude Code runs as an interactive session inside your terminal. Keep that session open while you add the marketplace.

Switch back to the terminal from earlier.
Start Claude Code by running this command:

claude

What does this command do?

The command starts an interactive Claude Code session in your current terminal. Slash commands entered next control Claude Code itself.

You'll see the Claude Code prompt waiting for your input.

Claude Code not starting?

Confirm that you ran the command in your normal terminal. A shell that cannot find Claude Code may not have the installation location on its command path.

Restart the terminal after checking the installation. Help me troubleshoot why Claude Code does not start from my terminal.

Add the treg marketplace by entering this command at the Claude Code prompt:

/plugin marketplace add superdesigndev/treg

What does this command do?

The marketplace command registers the official treg repository as a plugin source.
Claude Code can now discover the plugins published by that source.

Claude Code confirms that the treg marketplace is available.

Marketplace not appearing?

Check that the repository name is spelled exactly as shown in the command. Confirm that your computer can reach GitHub before trying again.

A managed Claude Code installation may block third-party marketplaces. Help me check why the treg marketplace cannot be added.

Open the treg plugin details by entering this command at the Claude Code prompt:

/plugin install treg@treg

What does this command do?

The command opens the treg plugin's details from the marketplace you just added. The details panel lets you review the components before choosing an installation scope.

Review the components listed in the plugin details.
Select Install for you to make treg available across your Claude Code projects.

Claude Code reports that the treg plugin is installed. That gives your agent the integration layer it needs for treg tools.

Plugin installation not completing?

Return to the plugin details and confirm that you selected an installation scope. Check the Errors tab if Claude Code reports a loading problem.

Refresh the marketplace if the plugin details are unavailable. Help me diagnose why treg@treg is not installing.

Connect treg's tools

The plugin gives Claude Code the treg integration. An MCP connection authorizes the agent to use treg's tools through your account.

This sign-in opens treg's authentication flow. Your token does not need to be pasted into the conversation.

Open Claude Code's connection manager by entering this command:

/mcp

What does this command show?

The connection manager lists the MCP services available to Claude Code. It also shows whether each service still needs authentication.

Select treg from the connection list.
Select Authenticate.
Complete the treg sign-in flow in your browser.
Return to Claude Code after the browser confirms the connection.

The treg entry now shows a connected state. The hard part is complete because Claude Code can now reach treg without exposing your account token.

treg still needs authentication?

Reopen the connection manager and select treg again. Complete the browser flow with the same treg account you used for the direct call.

Restart Claude Code if the browser confirms access but the connection state does not update. Help me troubleshoot the treg MCP connection.

Install and verify rea-growth

The rea-growth workflow is an opt-in skill. Its instructions guide the agent through collecting company growth signals and turning them into a blueprint.

Open the plugin manager by entering this command at the Claude Code prompt:

/plugin

What does this command show?

The plugin manager opens on the Discover tab. This tab lists plugins from every marketplace you have added.

Select the Discover tab.
Enter rea-growth into the plugin filter.
Select rea-growth from the filtered results.
Confirm that the plugin details list a skill.
Select Install for you.

Claude Code confirms that rea-growth is installed. Your agent now has the growth-research workflow alongside its treg connection.

Before you check the available skills, do you expect rea-growth to appear as a loaded skill or as an uninstalled marketplace result?

Open Claude Code's available skills list by entering this command:

/skills

What does this command show?

The skills list combines available skills from project sources and user sources. It also includes skills loaded from plugins.

You'll see rea-growth in the available skills list. You now have a connected agent that can follow the complete growth-teardown workflow.

rea-growth missing from the list?

Return to the Installed tab in the plugin manager. Confirm that rea-growth is enabled.

Restart Claude Code if the installation summary indicated that a reload was needed. Help me find why rea-growth is missing from the skills list.

Your agent can now call treg tools and follow the rea-growth workflow. Next, you'll give it a company and a time window for the teardown.

Generate a Growth Teardown

Your earlier setup connected Claude Code to treg. It also made the rea-growth skill available.

Your direct backlink call proved that one paid tool can return useful data. A credible growth teardown needs traffic, advertising, backlink, and social signals in one recorded investigation.

In this step, get ready to:

Define the teardown scope.
Run the rea-growth analysis.
Verify the growth-teardown run file.

Define the teardown scope

This analysis spends your remaining prepaid balance, so a small company gives you more control over the next set of calls. A defined time window keeps the investigation focused.

Select a small company for the teardown.
Record your choice here: selected company.
Choose a start date for the analysis.
Choose an end date for the analysis.
Record both dates here: defined time window.

Run the rea-growth analysis

The skill coordinates several treg calls to build a broader picture of the company's growth. Claude Code may pause for permission before some tool calls.

Before you submit the request, predict which signal group will produce the clearest growth clue.

Switch back to the Claude Code session from earlier.
Send the analysis request by pasting this prompt:

Use the rea-growth skill to analyze [[SELECTED_COMPANY="selected company"]] over [[TIME_WINDOW="defined time window"]].
Collect traffic, ads, backlink, and social signals through treg.
Write a growth-teardown run file that records the collected signals.
When the analysis is complete, confirm the run file path.
Summarize which signal categories were collected.

What does this request do?

The rea-growth skill coordinates paid data calls across the defined scope.
The signal list makes the agent gather evidence from four distinct growth channels.
The run-file instruction preserves the collected evidence for the blueprint.
Third-party traffic and advertising figures remain estimates.

Approve each treg tool call if Claude Code requests permission.
Keep the Claude Code session running until it returns a completion message.

You'll see Claude Code invoke treg tools across the requested categories. The run ends with a completion message that reports what the agent collected.

Analysis not starting?

Confirm that rea-growth remains visible in the available skills list.
Approve any pending treg tool permission in Claude Code.
Confirm that your treg account still has prepaid balance.
Help me troubleshoot why Claude Code is not starting the rea-growth analysis.

Verify the run file

The completion message is your checkpoint for this step. It should identify the saved run file and confirm the breadth of the collected evidence.

Before you inspect the result, predict whether the agent covered every requested signal group.

Wait for Claude Code to finish the analysis.
Review the final completion message.
Find the reported run file path.
Record the path here: growth teardown run file path.
Confirm that the completion message reports traffic signals.
Confirm that the completion message reports ads signals.
Confirm that the completion message reports backlink signals.
Confirm that the completion message reports social signals.

You should see a run file path. You should also see all four requested signal groups recorded as collected.

Run file confirmation incomplete?

Confirm that the final response says file writing completed.
Check that your submitted prompt requested a growth-teardown run file.
Help me locate and verify the growth-teardown run file.

That is the hard part done: your agent has turned separate paid data calls into one recorded growth investigation. Next, you'll open the finished blueprint and trace what each kind of call cost.

Review Your Growth Blueprint

Your analysis in Claude Code has turned the selected company plus time window into a growth teardown run file. This step turns that evidence into a growth blueprint you can use.

Third-party traffic figures can be estimates. This review checks the evidence plus metered costs before you choose the plays worth trying.

In this step, get ready to:

Open the completed growth blueprint.
Audit its evidence plus call costs.
Add a note naming the plays you would try.

Open the completed blueprint

A Markdown blueprint keeps the analysis readable outside the agent session. Its sections make missing evidence easier to spot.

Switch back to the Claude Code session from earlier.
Locate the completion message that reports the blueprint file path.
Use the reported file path to open the finished blueprint.

What Makes the Blueprint Complete?

Traffic sources capture where visits came from.
Ads capture paid acquisition signals.
Backlinks capture links from other sites.
Social signals capture activity across social platforms.
Copyable growth plays turn the findings into candidate actions.
The cost breakdown assigns spending to each call category.

Confirm the blueprint names the company you selected.
Confirm the blueprint uses the time window you defined.
Check that every section from the blueprint checklist appears in the file.

That first check is done: the finished blueprint is open with all required sections in view.

Missing a Blueprint Section?

Return to the same Claude Code session.
Ask the agent to identify the completed blueprint from the growth teardown run.
Compare the reported file against the blueprint checklist before continuing.

Need help tracing the output? Help me locate the completed growth blueprint from my Claude Code growth teardown run.

Audit the evidence and costs

Traffic plus advertising data from third-party sources can be estimates. Labels plus caveats show how much confidence to place in each claim.

Review the traffic sources section for estimate labels or caveats.
Review the ads section for estimate labels or caveats.
Compare the backlink evidence with the growth teardown run file from earlier.
Compare the social signals with the growth teardown run file from earlier.
Return to the cost breakdown.
Review the cost recorded for each call category.
Identify the category with the highest recorded cost.

You can now see which claims rely on estimates. You can also see which call category consumed the largest share of the teardown cost.

Choose the plays you would try

A recommended play becomes actionable once it fits your own goals. Your note records that decision beside the generated analysis.

Return to the growth plays section.
Review the copyable prompt attached to each play.
Choose the plays you would actually try.
Add a short note to the open Markdown blueprint naming your chosen plays.
Save the completed Markdown blueprint.

Before the final check, which part of the saved blueprint do you expect to reflect your own judgment?

Scan the saved blueprint from top to bottom.
Confirm all four signal categories are present.
Confirm the cost breakdown lists a cost for every call category.
Confirm your short note names the plays you would actually try.

The saved file should show four evidence categories: traffic sources, ads, backlinks, and social signals.

It should also show copyable growth plays plus costs by call category. Your short note names the plays you would actually try.

That is a strong finish: your growth blueprint now combines the agent's research with your own decision.

Can't See Your Note in the Saved File?

Confirm you saved the Markdown file that contains the cost breakdown.
Place the note inside that same file.
Repeat the final scan after saving.

Still missing something? Help me check whether my growth blueprint contains the required evidence, costs, plays, and personal note.

Secret mission

Audit a Growth Play

Audit one recommended growth play against the traffic, advertising, backlink, and social evidence in your blueprint. You will turn the recommendation into a focused experiment decision without making another paid call.

Clean Up Your Resources

Choose whether to keep your teardown files, pause your work for later, or delete the local outputs. Nothing recurs on your treg account unless you top up.

Your final balance records the cost of the direct backlink call and the rea-growth run.

Check your remaining prepaid balance by running this command:

treg balance

What should I see?

You should see your account and current prepaid balance. The amount reflects the calls made from this prepaid balance.

Balance Check Not Working?

Return to the terminal session where you authenticated with treg earlier.
Repeat the login flow from Step 1 if the session has expired.
Ask for help diagnosing the balance check with this troubleshooting prompt.

That final check closes the cost loop for your teardown.

Resources you used:

A local Growth Teardown Run File created by Claude Code with the collected traffic, advertising, backlink, and social signals.
A local Markdown Growth Blueprint containing the copyable plays, cost breakdown, and your chosen-play note.

Keep everything running

No action is needed while you want to keep your research available. The two files remain ready for future reference.

Keep the Growth Teardown Run File in the folder where Claude Code created it.
Keep the Markdown Growth Blueprint in the same folder.
Leave the treg CLI, treg plugin, and rea-growth skill installed for your next analysis.

The files stay available for reference. No recurring charge begins while the installed tools sit idle.

Pause - I'll come back to this later

Pausing means ending the active agent session while keeping the two files. Your prepaid treg balance stays available for a later run.

Save any open changes to the Markdown Growth Blueprint.
End the current Claude Code session.
Leave the two local files in their current folder.

Your teardown stays intact. You can return to the same files when you are ready to continue.

Delete - I don't want to use this again

Deleting the local outputs is permanent after your computer clears them. Your treg account does not need deletion to avoid recurring charges.

Move the Growth Teardown Run File to your computer's deleted-items location.
Move the Markdown Growth Blueprint to your computer's deleted-items location.
Confirm that neither file appears in the folder where Claude Code created it.
Clear the deleted-items location only when you are certain you do not need either file.

This removes both teardown files. The treg CLI, plugin, skill, and account remain available.

Nice Work!

You did it! You used Claude Code with the treg plugin plus the rea-growth skill to produce a Markdown growth blueprint for a real company.

You've learned how to:

Compare backlink-data tools by per-call price before choosing a data source.
Make a metered API call that returns raw JSON. Confirm its cost against the updated prepaid balance.
Create an evidence-backed growth blueprint from traffic signals across several channels. Evaluate the recommended plays by recording which ones you would actually test.
Secret Mission: Take on an optional challenge to push your growth-research skills further.

Ready to quiz yourself?