A practical guide to Indevaro MCP

Give your agent your app’s context.

Connect Claude Code or Codex to Indevaro. Read the evidence behind your app, choose a useful next step and prepare work you can review.

Set up Claude Code Set up Codex Try the prompts

By the Indevaro team. Reviewed 24 October 2026. Sources and verification

MCP, the Model Context Protocol, lets an AI client call tools from another service. Indevaro supplies tools that read your app’s context and, when you allow it, save work inside your workspace.

Your agent can start with the information you already have in Indevaro. You can ask it to explain a change, compare the available evidence or prepare a listing draft without pasting dashboard exports into every conversation.

Your AI clientClaude Code or Codex

Indevaro MCPTools you authorize

Your app’s contextAvailable data and saved work

Step 1

Start with an app in Indevaro.

An Indevaro account with an app you can access.
Claude Code or the Codex CLI installed and available in your terminal.
The relevant data connected in Indevaro. A public listing does not supply private revenue or product analytics.

You can begin on Free with its monthly allowance of read-only MCP calls. Saving changes through MCP requires a plan that allows writes and your permission in AI agent settings. Your AI client’s subscription or API costs are separate. Check current plans and limits.

Open AI agent settings to see connection instructions, connected agents and the account’s write setting.

Step 2

Connect the client you use.

The remote server address is https://app.indevaro.com/api/mcp. It uses Streamable HTTP and browser sign-in through OAuth. There is no Indevaro API key to paste into your project.

Claude Code

Run this in your terminal:

Claude Code setup command

claude mcp add --transport http indevaro https://app.indevaro.com/api/mcp

Open Claude Code in the project where you added the server.
Run /mcp, select Indevaro and choose Authenticate.
Complete sign-in in the browser. Review the agent name and return destination on Indevaro’s consent screen, then approve the connection if they match your client.
Return to Claude Code and confirm the server is connected.

The command uses Claude Code’s default local scope. Add it from the project where you will use it. See Claude Code’s MCP guide for other configuration scopes.

Codex

Add the remote server and start browser sign-in from your terminal:

Codex setup commands

codex mcp add indevaro --url https://app.indevaro.com/api/mcp
codex mcp login indevaro

Complete the browser sign-in using your Indevaro account.
Review and approve the connection on Indevaro’s consent screen.
Return to Codex. Start a new session if it was already open when you added the server.

Codex stores the server in ~/.codex/config.toml. Its CLI and IDE extension share that configuration. See OpenAI’s Codex MCP guide for client settings.

The terminal commands belong in a shell; the prompts below belong in your AI conversation. Cursor, VS Code and other clients have their own setup instructions in Indevaro’s AI agent settings.

Step 3

Ask about one app, with the evidence visible.

First, confirm the agent can see the right app. Keep this first session read-only.

Prompt 1: choose your app

Use Indevaro’s list_apps tool to list the apps I can access. Show their names and platforms, then ask which app I want to review. Do not create or change anything.

Choose an app from the response. Replace [APP NAME] with its name in the next prompt. Let the agent use the ID returned by the tool.

Prompt 2: read the context

Review [APP NAME] in Indevaro. Use its ID from list_apps and read get_overview. Summarize the available data, keeping sources, markets and reporting periods visible. Point out missing or stale information. Do not treat a provider-project total as revenue for this app. Do not refresh data or make changes.

Ask for a next step based on what is already available. A useful answer makes its source and uncertainty clear.

Prompt 3: choose a next step

For the same app, read the existing growth priorities and open cases. Suggest one useful next step, explain which evidence supports it, and tell me what we still need to verify. If there is no current priority or enough evidence, say so and propose a specific check. Keep the proposal in this chat; do not save or change anything in Indevaro.

Find the app: list_apps returns the apps the connection can access.
Read its context: get_overview gathers the available app overview. Source coverage and reporting periods still matter.
Pick up existing work: get_growth_priorities and get_open_cases read existing priorities and cases. An empty result is a reason to check the available context.

These are example prompts, not a fixed response transcript. Your answer depends on the chosen app, connected sources, available history and the model you use. Ask the agent to distinguish a tool’s observations from its own suggestions.

A real read: Haven, 24 October 2026

In our authenticated overview read, one keyword measurement was dated 24 October while the stored priority brief was marked expired. The App Store Connect connection was marked stale. These sources had different reporting times.

The useful next step is to check that source context in Indevaro before acting on the old priority. A successful tool response means the read worked; it does not make every measurement current. The read itself did not refresh those sources.

For an example of interpreting real measurements, read Haven’s weekly app review. The same questions apply here: which period, which population, what changed and what still needs checking?

Step 4

Turn the context into something you can use.

A listing description is a straightforward first task. Replace the app name and locale, then ask the agent to work from the existing listing:

Prompt 4: draft in the conversation

Read the current store listing for [APP NAME] and [LOCALE] from Indevaro. Draft one improved description here in the chat. Use only features supported by the listing or information I have confirmed. Explain each proposed change and list any questions. Do not save the draft or publish anything.

Read the proposal, correct unsupported claims and decide what to keep. Drafting text in your conversation does not save it in Indevaro.

When you want to save the work

Make the destination explicit.

On a plan that allows MCP writes, enable Allow agents to make changes in AI agent settings. Then ask the agent to save only the text you have reviewed, for the selected app and locale, using save_listing_draft. It should use the latest listing snapshot and revision, preserving the fields you did not change.

This saves a draft inside Indevaro. It does not update the live App Store listing. Review the saved draft in Indevaro before taking a separate publication step.

For work attached to an existing case, prepare_action can prepare an Action draft. Its prerequisites and approval still apply; a prepared draft is not evidence that the action was approved, executed or successful.

You decide what the agent can change.

Read access: the connected agent can request data available to your account. Consider which AI client you want to receive that information.
Write access: Free is read-only. Where your plan permits writes, the account’s write switch must also be enabled. Your client may ask for approval before individual tool calls.
Saved work: listing and Action drafts stay inside Indevaro. MCP does not publish to App Store Connect, alter Apple Ads campaigns or connect new integrations.
Revocation: open Account settings → AI agents → Connected agents, choose Revoke, and confirm Revoke access. The agent must be approved again to reconnect.

A prompt asking the agent not to write is useful guidance. The plan restriction and account write setting enforce the permission boundary.

If something gets stuck

The client has no Indevaro tools.

Check the server address, confirm authentication completed and reopen the client session. In Claude Code, check /mcp. In Codex, check codex mcp list and run codex mcp login indevaro again if sign-in is missing.

I cannot see my app, or some data is missing.

Confirm you signed in with the account that owns the app. Open its workspace in Indevaro and check the relevant connection, coverage and latest reporting period. MCP reads the available context; connecting an agent does not connect RevenueCat, store accounts or product analytics for you.

The agent can read, but cannot save a draft.

Check your plan and Allow agents to make changes in AI agent settings. A monthly call limit, paused app or a tool’s missing prerequisites can also stop work. Read the returned error and resolve it in Indevaro instead of repeatedly retrying.

The sign-in page shows a different destination than I expected.

Return to the client you intended to connect and check its configuration. Approve only a connection you recognize. A loopback address such as localhost can be the client’s browser callback; it is not the Indevaro server address you should configure.

Can I use this without writing any code?

Yes. After the one-time client setup, the prompts are ordinary requests in your AI conversation. The guide uses the terminal to configure Claude Code or Codex; it does not require changes to your app’s source code.

Sources and verification

Reviewed on 24 October 2026 against Indevaro’s client configuration and tool permissions, the client documentation below and Codex CLI help. An authenticated list_apps → get_overview read was checked against a real workspace. We did not repeat a fresh OAuth sign-in for every client or save production data while preparing this guide.

Commands follow each client’s supported setup. Client interfaces and account policies can change; use the vendor documentation if your installed version differs.

Start with one question about your app.

Connect your agent, choose an app and ask what the available evidence supports. Keep the first review read-only, then decide which work to save.

Open AI agent settings

An Indevaro account is required. Sign in if prompted.