Stripe launches agent payments

Plus: LangChain's State of AI Agents report, and an review of the companies making an impact in the agent space

Author

Michael Cunningham
November 18, 2024

In partnership with

Welcome back to Building AI Agents, your biweekly guide to everything new in the AI agent field!

Who’s winning the great Salesforce-Microsoft AI agent war? Meme makers

In today’s issue…

  • Stripe joins the race to build agent payment platforms

  • LangChain releases its State of AI Agents report

  • A roundup of the major agent software providers

  • The 6 most impactful business use cases for agents

…and more

🔍 SPOTLIGHT

Created by the authors

Autonomous AI is ready to give humans payback—though the reality is less menacing than it sounds.

On Thursday, payment processor Stripe announced the launch of its Stripe Agent Toolkit, a software development kit (SDK) which enables AI agents to engage in transactions with human users and to pay third parties, such as hotels or airlines. With the company’s Stripe Issuing technology, agents can create single-use virtual credit cards that can be employed to pay for products with a simple LLM function call. The toolkit integrates with common agent platforms such as CrewAI, LangChain, and Vercel’s TypeScript SDK, allowing agent builders to easily slot it into existing architectures.

Though Stripe is by far the highest-profile company to facilitate agent payments, it is by no means the first—the past month has seen a surge of launches by startups aiming to capitalize on the new AI agent economy.

In early October, we reported on Payman, a pre-seed company which bills itself as the first AI-to-human payment platform and is intended to allow AI agents to hire human workers—a likely use case being to pay freelancers to accomplish tasks that are beyond an agent’s current capabilities.

Startup Skyfire, meanwhile, takes a crypto-centric approach, offering a platform in which agents can send and receive payments in USD Coin, a cryptocurrency pegged to the US dollar, and emphasizes verification of an agent’s identity in order to establish trust with transaction partners. The company recently secured funding from Andreessen Horowitz’s cryptocurrency startup accelerator CSX and crypto exchange Coinbase’s Ventures fund, bringing its total capital raise to $9.5 million. The cryptocurrency world has been rapidly embracing AI agents, with Coinbase itself rolling out its Coinbase Developer Platform to facilitate agentic payments.

With voice agents capable of engaging in fluid conversations over the phone on the rise, YC-backed startup Protegee aims to enable them to accept payments from human buyers. In a typical scenario, a human could call a hotel to book a room, work with its AI voice agent to identify a suitable one, then pay the agent through the phone using their credit card via Protegee’s tech.

As AI agents become more capable, the range of uses they are being entrusted with is progressively expanding to include activities with real-world economic significance. The recent wave of agent payment startups is a reflection of this reality, presaging a future in which agents are not just used to automate rote business and personal tasks, but become increasingly autonomous commercial actors in their own right.

If you find Building AI Agents valuable, forward this email to a friend or colleague!

Refer one person and we’ll send you a set of weekly bonus links to additional AI agent news for a month! See here for a sample

🤝 WITH WRITER

Writer RAG tool: build production-ready RAG apps in minutes

  • Writer RAG Tool: build production-ready RAG apps in minutes with simple API calls.

  • Knowledge Graph integration for intelligent data retrieval and AI-powered interactions.

  • Streamlined full-stack platform eliminates complex setups for scalable, accurate AI workflows.

📰 NEWS

Source: LangChain

The company laid out its findings from a survey of over 1,300 professionals on their use of AI agents, revealing that—among other striking takeaways—non-tech companies are as likely as tech companies to be making use of them.

In a somewhat confusing split, the former AutoGen team led by Chi Wang has completed its fork from Microsoft, renaming its version AG2. Microsoft’s own fork of the multi-agent framework will continue to be maintained and will inherit the AutoGen name.

NEO is a multi-agent ML engineering system which scored a medal in 26% of the 50 Kaggle competitions it was tested against, significantly outperforming the previous state-of-the-art 16.9%.

🛠️ USEFUL STUFF

Source: Letta

Agent startup Letta presents a comprehensive overview of the major providers in the agent space, split across niches such as model serving, storage, agent frameworks, and more.

screengrasp is a click positioning model which is claimed to outperform Anthropic Computer use, OmniParser, and Molmo for correctly clicking elements on a screen, a crucial task for visual computer use agents.

TinyTroupe is a multi-agent framework for creating teams of AI agents with distinct personas for roleplaying social situations. Unlike other agent platforms, it is intended to enable realistic simulation of human opinion and interaction for applications such as evaluating ad campaigns and products.

💡 ANALYSIS

Source: CIO

Based on interviews with experts in the space, the author of this piece gives an outline of the 6 main business use cases AI agents will be put to, from software development to business intelligence and more.

Consulting firm EY discusses the roles that agents will fulfill—and how business leaders should prepare for them.

This piece by Bain Capital argues that AI agents will usher in an “Agentic Commerce Era” powered by shopping agents which will be empowered to pick out products on behalf of consumers.

🧪 RESEARCH

Source: ArXiv

This paper explores the major challenges inherent in deploying LLM agents, such as observability and traceability, and identifies essential tools which address them.

AI agents tend to perform badly on game theory tasks, deviating from optimal strategies. The authors of this paper propose game-theoretic workflows which allow agents to effectively make more rational choices, with significant implications for tools such as negotiation agents.

Thanks for reading! Until next time, keep learning and building!

What did you think of today's issue?

Login or Subscribe to participate in polls.

If you have any specific feedback, just reply to this email—we’d love to hear from you

Follow us on X (Twitter), LinkedIn, and Instagram