Agent-installed iOS feedback
Describe the feedback moment. Let your coding agent wire it into iOS.
Paydirt gives Codex, Claude Code, and other MCP-capable coding agents a concrete installation contract: create the form, edit the host app, connect Slack, build, and return a test path.
What this workflow actually does
Developer tools often call a task complete after printing a package URL. Paydirt’s agent workflow defines a stricter finish line. The agent first inspects the iOS project, detects the dependency manager, existing feedback behavior, and subscription source of truth, authenticates through the MCP flow, creates or reuses the requested forms, edits only the requested triggers, assigns Slack, and builds the target.
A product request can stay natural: “Add a feedback form titled Export Feedback after a successful export.” The placement tool returns an idempotent form and an explicit trigger contract, but the coding agent still owns locating the successful export path and compiling the change.
The agent installs the form types the user requests. Existing feedback UI is preserved, and regular feedback, trial cancellation, subscription cancellation, and named contextual forms remain equally supported.
The MCP also makes completed responses readable by coding agents. That access is informational in version 1: Paydirt does not automatically open issues, write code, or act on the customer’s words.
A strong fit for
- ✓ Developers using Codex, Claude Code, or another MCP host
- ✓ Existing iOS apps where the correct trigger must be discovered in code
- ✓ Teams that want repeatable RevenueCat, Superwall, StoreKit, or manual setup
- ✓ A build-and-test handoff instead of generic installation instructions
Not designed to replace
- — Agents without repository write access or an iOS build environment
- — No-review autonomous changes from customer feedback
- — Replacing product judgment about where a form belongs
- — Pretending a remote form is installed before app code compiles
From trigger to exact answers
01
Ask in product language
Name the feedback use case, desired title, in-app location or action, and Slack destination.
02
Inspect before installing
The agent discovers bundle ID, targets, package style, app entry point, product IDs, and purchase provider.
03
Use the Paydirt tools
`paydirt_setup` returns forms and provider-specific actions; `paydirt_add_feedback_form` handles named placement requests.
04
Prove the host app
The agent edits the exact trigger, connects every form to Slack, builds the real target, and gives a reproducible test path.
Register once, then ask naturally
The MCP package is public and versioned on npm. Registration makes Paydirt’s setup and form-placement contracts available in the coding agent.
# Codex
codex mcp add paydirt -- npx -y paydirt-mcp@latest
# Claude Code
claude mcp add paydirt -- npx -y paydirt-mcp@latest
# Then ask:
Add a feedback form titled "Export Feedback" after a successful export,
send completed conversations to #product-feedback, build, and test it.What reaches Slack
- • A durable remote form tied to the requested app
- • A concrete Swift trigger in the host repository
- • Slack assignment for every installed form
- • A successful app-target build and exact manual test steps
- • Read-only access to completed raw responses for later analysis
Know the limits
- • The MCP host may require a reload after first registration
- • OAuth still requires the human to authorize Paydirt and Slack in a browser
- • The coding agent needs permission to edit and build the iOS project
- • An agent can implement the request, but the developer remains responsible for reviewing UX, privacy, and release behavior
Questions developers ask
Can I just say “Add a feedback form titled XYZ at this action”?
Yes. The agent calls `paydirt_add_feedback_form`, creates or reuses the named form, edits the exact host-app action, assigns Slack, builds, and reports a test path.
What happens to my existing feedback form?
It stays unchanged unless you explicitly ask the agent to modify it. Paydirt forms are added at the requested feedback moments while preserving the app’s existing feedback behavior.
Does the agent need to know whether the app uses RevenueCat?
No. It should inspect the project and select RevenueCat, standalone Superwall, native StoreKit, or a manual app-owned event. You should confirm only when the source of truth is genuinely ambiguous.
Will Paydirt automatically change code based on responses?
No. Responses are information for Slack and coding agents. Any later action requires a separate human request.
Ship the smallest useful feedback loop
Install one form at a real product moment, send the completed raw conversation to one Slack channel, and verify it in the app before expanding.
Start setup