August 5, 2025
5 min read
Decrypt / Jason Nelson
OpenLedger’s crypto AI agent requires user approval before trading, ensuring transparency and control in a fast-paced market.
In a Market Obsessed With Speed, This Crypto AI Agent Wants You to Slow Down
While most crypto bots rush into trades without a second thought, San Francisco–based OpenLedger is slamming on the brakes. Its new AI agent won’t execute a single transaction until the user explicitly signs off, bringing humans back into the loop of automated trading. That matters, the company argues, because speed without context can spell disaster. OpenLedger’s approach balances rapid execution with real-time adaptability and full transparency. Traders set their risk limits and review every recommendation before it goes live, ensuring the AI never becomes a black box. “People need to understand why a trade is happening, not just watch their balance change,” OpenLedger core contributor Kamesh told Decrypt. Once launched, OpenLedger’s agent will live inside Trust Wallet and communicate in plain English. It can summarize positions, flag unusual activity, and suggest strategies. According to Kamesh, while AI can be a powerful tool, it shouldn’t remove humans from decisions. “The AI doesn’t need signing authority to be useful,” he said. “Think of it like a smart co-pilot—you’re still flying the plane.” That restraint is by design, Kamesh said. Users define their parameters—risk thresholds, network preferences, and data-sharing settings—before the agent ever makes a move. All prompts and context stay on the device by default, so private keys and transaction histories never leave your phone. To keep every decision traceable, OpenLedger uses a Proof of Attribution framework. Each suggested action links back to its data sources and AI model version, so users can audit why a trade was recommended. If the data trail is incomplete, the agent simply won’t proceed. “We prioritize privacy, so nothing is stored in the cloud unless explicitly needed for functionality, and even then, it's anonymized and never linked to your wallet or identity,” Kamesh said. OpenLedger will introduce the agent in Trust Wallet in October, starting with a pilot for a few hundred users. Wider access will follow early next year. The OpenLedger AI launch comes amid a boom in crypto of AI-powered tools. The sector’s market cap hit $5.7 billion in July, according to CoinGecko, as developers race to bring AI onto DeFi platforms. OpenLedger isn’t alone in bringing AI to wallets. In July, Base, the newly rebranded wallet app by Coinbase, announced an AI feature that lets users trade and send crypto using natural language prompts. The rush to add AI to crypto transactions has, however, also raised concerns about “hallucinations,” or inaccurate outputs that can lead to financial losses. In May, researchers at Princeton University showed how memory-based manipulation attacks could trick crypto AI agents into approving unauthorized transactions. OpenLedger says its system is built to resist that kind of exploit. “You don’t have to trust what the AI says blindly,” Kamesh said. “You can trace the logic, the data behind its decision, and even who contributed that data. If the source can’t be traced, the agent doesn’t act.”Frequently Asked Questions (FAQ)
OpenLedger AI Agent Features
Q: What is the primary goal of OpenLedger's new AI agent? A: The primary goal is to bring a more controlled and human-in-the-loop approach to automated crypto trading, emphasizing user oversight and understanding over raw speed. Q: How does OpenLedger's AI agent differ from other crypto trading bots? A: Unlike many bots that execute trades automatically, OpenLedger's agent requires explicit user approval for every transaction, ensuring human control and transparency. Q: In which wallet will OpenLedger's AI agent be available? A: The AI agent will be integrated within Trust Wallet. Q: What kind of functionalities will the AI agent offer? A: It will be able to summarize positions, flag unusual activity, and suggest trading strategies, all communicated in plain English.Transparency and Security
Q: How does OpenLedger ensure transparency in its AI agent's recommendations? A: OpenLedger uses a Proof of Attribution framework, linking each suggested action to its data sources and AI model version for user auditing. Q: What measures are in place to prevent the AI agent from becoming a "black box"? A: Users set their own risk limits and review all recommendations before they are executed, ensuring they understand the reasoning behind each proposed trade. Q: How does OpenLedger handle user privacy and data security? A: Prompts and context remain on the user's device by default. Private keys and transaction histories do not leave the phone, and cloud storage is only used for anonymized data when explicitly needed for functionality. Q: What is the Proof of Attribution framework and how does it enhance security? A: This framework makes every suggested action traceable by linking it to its data sources and AI model version. If the data trail is incomplete, the agent will not proceed, preventing actions based on untrusted information.AI in Crypto Trading
Q: Why is it important for humans to remain involved in AI-driven crypto trading? A: Speed without context can lead to disaster. Human oversight ensures understanding of trading decisions and mitigates risks associated with AI "hallucinations" or errors. Q: What are "hallucinations" in the context of AI crypto trading? A: Hallucinations refer to inaccurate outputs from AI that could lead to incorrect recommendations or unauthorized transactions. Q: How does OpenLedger's system address the risk of AI "hallucinations" or exploits? A: By requiring user sign-off for every transaction and providing a traceable logic for recommendations, users can audit decisions and avoid blindly trusting the AI. The Proof of Attribution framework also ensures that actions are only taken with traceable data sources.Originally published at Decrypt on August 4, 2025.