> ## Documentation Index
> Fetch the complete documentation index at: https://ifdocdemo.lumivescence.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Integrate with a coding agent

> Integrate InFlow Payments, Enrollment, and Discovery with a coding agent. It plans the integration with you and builds one approved step at a time.

A coding agent can build the three InFlow integration steps with you: Payments (MPP and x402), Enrollment (AEP), and Discovery (ODP). It turns your goals into a plan and builds one approved step at a time. See [Integrate your service](/sellers/integrate-your-service) for the steps.

Any coding agent that can read a URL and edit a repository works, for example Claude Code, Cursor, or Codex.

## Start in your repository

<Steps>
  <Step title="Open your repository">
    Open your application repository in your coding agent.
  </Step>

  <Step title="Paste the prompt">
    Paste this prompt into the chat of your coding agent.

    ```text theme={null}
    Read https://directory.inflowpay.ai/integration.md and follow its guided process to integrate Agentic Commerce into this repository. Pause at each stage and wait for my response.
    ```
  </Step>

  <Step title="Choose all three">
    The agent asks: "Would you like to integrate all three or a subset?" Choose all three. The agent does not choose for you.
  </Step>

  <Step title="Type go to continue">
    The agent confirms the repository and saves a coordination plan. Then it asks you to type `go`. Each `go` approves only the next step that the agent names.
  </Step>
</Steps>

## The order of the work

The agent builds **Payments, then Enrollment (AEP), then Discovery (ODP)**. This order is a recommendation, not a protocol rule. At runtime, an agent uses your service in the opposite order: it discovers with ODP, enrolls with AEP, and then pays.

Each step has its own guide and its own plan file in your repository.

| Step | Guide that the agent reads | Plan file |
| - | - | - |
| All three steps | `directory.inflowpay.ai/integration.md` | `COMMERCE-INTEGRATION-PLAN.md` |
| Payments | `app.inflowpay.ai/integration.md` | `INFLOW-INTEGRATION-PLAN.md` |
| Enrollment (AEP) | `www.aep.foundation/integration.md` | `AEP-INTEGRATION-PLAN.md` |
| Discovery (ODP) | `www.offeringprotocol.org/integration.md` | `ODP-INTEGRATION-PLAN.md` |

The coordination plan records your choices, the order, the current stage, and the next action. Each step guide does its own interview, repository review, SDK choice, plan, and checks.

## You stay in control

| Rule | Detail |
| - | - |
| You review the plan | The agent changes no code before you approve the plan. During the assessment, it writes only the plan files. It does not install packages, edit application code, or run paid requests. |
| You approve each step | `go` approves only the next step. |
| Some actions need separate approval | State-changing checks, enrollment, payments, deployment, publication, commits, and pull requests each need their own approval. |
| Payments need limits | Before a payment, the agent confirms the environment, the destination, the operation, and the spending limit with you. |
| No secrets in chat | The agent never puts credentials, keys, tokens, or personal data in chat or in a plan. |
| Your checks stay | Your tests and quality checks stay part of the work. |

## How the agent knows the work is done

Each step plan ends with two groups of exit criteria.

| Group | What it covers |
| - | - |
| Project acceptance | Your tests, coverage, static analysis, and regression checks. |
| InFlow acceptance | InFlow CLI checks against the running integration. For ODP, also the directory publication. See [Check your integration](/sellers/verify-your-integration). |

The agent keeps each state separate: local tests, checks on the running service, deployment, submission, and listing. "Implementation complete; deployed verification pending" does not mean complete. A successful submission does not prove that the directory indexed your service.

## Run the guide again

The agent saves the plans and the progress in your repository. You can pause and resume, or start a new assessment when your goals change. On a new run, you can choose one improvement, for example ODP from the Minimal [level](/sellers/make-products-discoverable#step-2-choose-a-level) to the Complete level.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.