Coinbase’s Base narrows its $100,000 startup accelerator to 10 AI-agent teams

Coinbase’s Base, the blockchain network tied to the U.S. crypto exchange, is preparing a new round of funding and mentorship designed around a single theme: AI agents. According to a report published by Decrypt (via Yahoo Finance), Base is narrowing its startup accelerator to 10 teams and offering $100,000 for participating startups as it searches for developers building products that combine agentic artificial intelligence with crypto-native use cases.

The accelerator focus, as described in the report, goes beyond pure AI. Base is specifically looking for teams working on AI agents that can operate in areas such as payments, trading, and financing. Those categories point to a broad effort to translate agent behavior into practical onchain workflows, from initiating transactions to managing trades and potentially supporting automated lending or credit-like functions.

For Base, the timing fits into a larger shift across crypto infrastructure toward AI-related product experiments. In recent years, blockchain networks have competed to attract developers with faster onboarding, better developer tooling, and distribution advantages. An accelerator that targets a narrow cohort can function as a concentrated “builder pipeline,” helping the network identify early projects that can demonstrate real usage rather than technology prototypes alone.

The program’s structure is also notable because of how small it is. Base is not described as running a large class of startups for this round, but instead selecting a limited number of teams, capped at 10. That approach can reduce noise and improve day-to-day interaction between mentors and founders, although it also means many applicants will be rejected and the funded pipeline will be thinner than broader accelerators.

What is clear from the reporting is the company’s selection criteria and funding headline: $100,000 and an accelerator cohort limited to 10 startups. What is not clear in the publicly described summary is the split of that money, such as whether the $100,000 is structured as grants, convertible instruments, equity investments, or a combination. The report also does not lay out how Base plans to evaluate the teams, beyond the general emphasis on AI agents and the listed product areas.

Neither the report summary nor the information provided here spells out the timeline for the accelerator, including when startups begin and when results are expected. It also does not specify whether Base intends to integrate the winning teams’ products into its own ecosystem through partnerships, funding follow-ons, or additional technical support. Without those details, observers will have to wait for an official program announcement to understand how Base’s incentives are designed to accelerate actual adoption.

For the crypto sector, the move underscores a growing belief that AI agents could become a new interface layer for decentralized finance and digital asset activity. If agents can reliably interpret instructions, execute actions, and manage multi-step processes, they could lower the barrier for consumers and institutions that currently find crypto tooling too complex or too manual. The key question is whether these experiments can meet safety and reliability expectations, especially when automation touches payments, trading, or financing.

Base’s accelerator round will also raise practical questions that the current reporting does not answer, particularly around guardrails. AI agents in financial contexts require careful controls, auditability, and clear permissions. Until Base discloses details about compliance expectations, security practices, and any required architecture, it will be difficult to gauge how the program intends to balance innovation with risk management. What to watch next is the formal program terms, selection criteria, and the list of the 10 startups chosen, which will likely offer the clearest announcement of which AI-agent use cases Base believes are ready for real-world deployment.