---
title: "Adding payments to your LLM agentic workflows"
description: "Agentic workflows combine large language models with function calling to turn prompts into sequences of actions, but connecting those actions to financial services requires controlled API access. The Stripe agent toolkit integrates Stripe with Vercel’s AI SDK, LangChain, and CrewAI, supports function-calling LLM providers, and is built on Stripe’s Node.js and Python SDKs. It exposes configurable tools for tasks such as invoicing, creating and sharing payment links, and using Issuing virtual cards with spending limits, authorization checks, and deactivation after purchase. Middleware can record input and completion token counts as billing events for usage-based customer billing. Because agent behavior is non-deterministic, the post recommends test mode, evaluations, restricted keys, limited tool sets, and reduced request and response shapes to improve focus and constrain failures."
---

# Adding payments to your LLM agentic workflows

[Stripe](https://yomu.fyi/company/stripe) · Steve Kaliski · Nov 14, 2024

**Type:** Explainer

## Summary

Agentic workflows combine large language models with function calling to turn prompts into sequences of actions, but connecting those actions to financial services requires controlled API access. The Stripe agent toolkit integrates Stripe with Vercel’s AI SDK, LangChain, and CrewAI, supports function-calling LLM providers, and is built on Stripe’s Node.js and Python SDKs. It exposes configurable tools for tasks such as invoicing, creating and sharing payment links, and using Issuing virtual cards with spending limits, authorization checks, and deactivation after purchase. Middleware can record input and completion token counts as billing events for usage-based customer billing. Because agent behavior is non-deterministic, the post recommends test mode, evaluations, restricted keys, limited tool sets, and reduced request and response shapes to improve focus and constrain failures.

## Context

Agentic workflows use non-deterministic model behavior to perform multi-step actions, including financial operations. The post notes that purchases should match user intent and that agent workflows have material costs typically measured by token use or time.

## Approach / What changed

The Stripe agent toolkit exposes selected Stripe functionality as tools for Vercel’s AI SDK, LangChain, CrewAI, and other function-calling frameworks. It also provides Issuing controls for virtual-card spending and middleware that tracks prompt and completion tokens for metered billing. Configuration can restrict API actions and reduce request and response shapes.

## Takeaways

- Stripe tools can be combined with other toolkits, such as Slack, to create multi-step workflows that create a payment link and post it to a channel.
- Issuing virtual cards can be limited by spending controls, while real-time authorization handling can approve or decline purchases and help verify that agent actions match user intent.
- The toolkit recommends test mode, evaluations, restricted API keys, narrowly selected actions, and reduced API payloads because agents are non-deterministic and can struggle with larger tool surfaces.

**Tags:** [Authentication](https://yomu.fyi/topic/authentication), [LLMs](https://yomu.fyi/topic/llm), [Python](https://yomu.fyi/topic/python), [Reliability](https://yomu.fyi/topic/reliability), [TypeScript](https://yomu.fyi/topic/typescript)

- Source: [Stripe](https://stripe.dev/blog/adding-payments-to-your-agentic-workflows)
- Source URL: https://stripe.dev/blog/adding-payments-to-your-agentic-workflows
- Ingested by Yomu: 2026-08-28T08:56:59.838Z

[Read original post](https://stripe.dev/blog/adding-payments-to-your-agentic-workflows)