# US 2026/0099301 A1 · Automated code generation using large language models for software platform integrations with computing services

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0099301 A1
- Document ID: US20260099301A1
- Status: Published application
- Kind: A1
- Application number: 18/911,167
- Filed: 2024-10-09
- Published: 2026-04-09
- Earliest claimed: 2024-10-09
- Assignees: Paypal, Inc.
- Applicants: Paypal, Inc.
- Inventors: Shivani Bayala Banerjee; Taehee Roy Han; Laxmikant Sharma; Chandra Sekhar Nallam; Mayank Dandotia; Mudita Tiwari
- Web page: https://patentgrep.com/patent/US20260099301A1

## CPC

- G06F 8/30: Creation or generation of source code
- G06F 8/10: Requirements analysis; Specification techniques
- G06F 8/65: Updates (security arrangements therefor G06F21/57)
- G06N 20/00: Machine learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

There are provided systems and methods for automated code generation using large language models (LLMs) for software platform integrations with computing services. An online transaction processor or other service provider may provide computing services and platforms to entities including merchants for electronic transaction processing and other account services. To provide for conde integrations of the computing services with software platforms of merchants, such as software systems for merchant websites and applications, the service provider may provide for an automated code generation tool that processes data for the software platform and the service to be integrated and provides recommendations and source code. An LLM may be prompted in parts to generate the code after analyzing legacy code from previous integrations and information for the merchant's software platform, and the merchant may be provided new code for testing and implementation.

## Claims

### Claim 1

A system comprising:
a non-transitory memory; and
one or more hardware processors coupled to the non-transitory memory and configured to execute instructions to cause the system to:
receive a request associated with integrating a software platform of a merchant with a computing service associated with a service provider, wherein the request comprises a payload associated with the software platform;
determine a large language model (LLM) prompt for an LLM configured to generate new code integrations with at least the computing service based on LLM training data associated with available code integrations for at least the computing service;
prompt the LLM to create a new code integration based on the LLM prompt and the payload;
receive the new code integration for the software platform from the LLM, wherein the new code integration comprises a code package of new code generated by the LLM for one or more uses of the computing service of the service provider via the software platform of the merchant; and
load the code package for the new code integration to a selected computing system designated by the merchant.

### Claim 2 (depends on 1)

The system of claim 1, wherein the request comprises an update of existing code with a set of legacy application programming interfaces (APIs) previously used with the computing service to the new code for a set of current APIs for the computing service.

### Claim 3 (depends on 2)

The system of claim 2, wherein the existing code comprises source code currently used by the software platform, wherein the payload comprises one or more code packages for the source code, and wherein the set of current APIs include one or more representational state transfer (REST) APIs.

### Claim 4 (depends on 1)

The system of claim 1, wherein executing the instructions further causes the system to:
determine a software design document (SSD) for the software platform,
wherein the prompting is further based on the SSD.

### Claim 5 (depends on 1)

The system of claim 1, wherein the LLM prompt comprises a plurality of LLM prompts, and wherein prompting the LLM is further based on each of the plurality of LLM prompts and a corresponding portion of the payload.

### Claim 6 (depends on 5)

The system of claim 5, wherein prompting the LLM uses an orchestration layer of the system to prompt the LLM using each of the plurality of LLM prompts and the corresponding portion of the payload, and wherein the plurality of LLM prompts are associated with generating the new code for at least one of an authentication process, backend code, a backend API layer, a user interface, one or more user interface elements, or one or more application dependencies.

### Claim 7 (depends on 1)

The system of claim 1, wherein the new code, when executed, integrates the computing service on the software platform of the merchant for use by one or more users via a website or an application of the merchant, and wherein the computing service comprises at least one of a user authentication service, a transaction processing service, or a digital account service.

### Claim 8 (depends on 1)

The system of claim 1, wherein loading the code package comprises pushing the code package to the selected computing system designated by the merchant, and wherein the selected computing system comprises one of a test computing environment, a production computing environment, or a code repository platform.

### Claim 9 (depends on 1)

The system of claim 1, wherein prompting the LLM is further based on sample code for at least one of the available code integrations.

### Claim 10

A method comprising:
determining that a code integration with a computing service provided by a service provider is available for a computing platform;
determining a payload usable when prompting a large language model (LLM) for an automated generation of the code integration, wherein the LLM is configured to generate new code integrations with at least the computing service based on LLM training data associated with available code integrations for at least the computing service;
identifying an LLM prompt associated with the code integration and the LLM;
prompting the LLM based on the LLM prompt and the payload;
receiving, based on the prompting, a new code integration for the computing platform from the LLM, wherein the new code integration comprises a code package of new code for one or more uses of the computing service of the service provider via the computing platform; and
outputting the new code integration via a code integration deployment process for the code integration of the computing platform with the computing service.

### Claim 11 (depends on 10)

The method of claim 10, wherein the LLM has a token limit corresponding to a data restriction for the prompting the LLM, and wherein the prompting the LLM is performed based on the token limit.

### Claim 12 (depends on 11)

The method of claim 11, wherein the prompting the LLM comprises prompting the LLM in parts based on a plurality of calls executed to the LLM to adhere to the token limit.

### Claim 13 (depends on 12)

The method of claim 12, wherein the prompting the LLM in parts uses different portions of the payload for each of the plurality of calls.

### Claim 14 (depends on 12)

The method of claim 12, wherein the prompting in parts is performed based on different code sections for generation by the LLM for the new code.

### Claim 15 (depends on 10)

The method of claim 10, wherein the payload comprises at least one of a legacy code integration of the computing service with the computing platform, a software design document (SSD) for the software platform, or one or more response payloads of API calls from the software platform.

### Claim 16 (depends on 10)

The method of claim 10, wherein the new code implements the computing service externally on the computing platform and causes calls to be executed to the service provider for the computing service.

### Claim 17 (depends on 10)

The method of claim 10, wherein outputting the new code integration comprises storing the new code with a code repository platform accessible by a plurality of users, and wherein the new code is testable and reviewable on the code repository platform.

### Claim 18 (depends on 10)

The method of claim 10, wherein the prompting the LLM is further based on sample code previously used by at least one of the available code integrations.

### Claim 19

A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:
receiving a request for a use of a computing service of a service provider with a software platform of a merchant, wherein the request comprises computing code data associated with the software platform;
determining at least one large language model (LLM) prompt for an LLM configured to generate a new code integration of the computing service with different software platforms based on LLM training data associated with integrating code for the computing service with the different software platforms;
prompting the LLM based on the at least one LLM prompt and the computing code data;
receiving the new code integration for the software platform from the LLM, wherein the new code integration comprises new code generated by the LLM for the use of the computing service of the service provider via the software platform of the merchant; and
communicating the code package to the merchant.

### Claim 20 (depends on 19)

The non-transitory machine-readable medium of claim 19, wherein the at least one LLM prompt is associated with a plurality of computing code tokens generated based on a maximum data size for LLM tokens processable by the LLM, and wherein the plurality of computing code tokens correspond to different code portions of the new code integration and corresponding data portions of the computing code data.
