發生了什麼事
IT Brief Australia reported that Olas launched Pearl Connect, a beta product for Claude Code and Claude Desktop users on the Gnosis and Polygon networks. The service runs locally within Olas’s Pearl app ecosystem and lets coding agents check balances, request marketplace services and prepare blockchain transactions, while users retain control of signing and private keys.
IT Brief Australia reported that Olas launched Pearl Connect in beta for users of Claude Code and Claude Desktop. According to the report, the product is part of Pearl, Olas’s AI-agent app store, and provides a self-custodial crypto wallet that can be used during a coding session. The initial supported networks are Polygon and Gnosis. The report said the service is aimed first at prediction markets, allowing a user to research a market, obtain a probability estimate from a specialised agent and prepare an on-chain transaction without moving to a separate wallet or browser.
The report described Pearl Connect as a user-directed co-pilot rather than an autonomous trading system. IT Brief Australia distinguished it from Polystrat, another Pearl product that the outlet said trades continuously on Polymarket without manual input. With Pearl Connect, the coding agent can check wallet balances, request services from the Olas Marketplace and prepare transactions for applications including Polymarket and Omen. IT Brief Australia reported that each Connect agent operates on one chain, with its own wallet and funding, and that users can create additional agents for other supported chains.
IT Brief Australia reported that Pearl Connect creates a Safe smart account and runs locally within Pearl. The outlet said transactions are signed locally on the user’s machine, with private keys kept outside the AI session. It also reported that the wallet remains owned by the user and can be restored through a Pearl password, recovery phrase or backup wallet. The service reportedly supports interactions with DeFi protocols and other compatible smart contracts specified by the user, although Olas positioned prediction-market activity as the beta’s main entry point. The source does not include independent confirmation from Olas documentation, Claude’s developers, the supported networks or transaction records.
The report also described an agent-to-agent payment model. IT Brief Australia said a coding agent could pay a specialised prediction agent on the Olas Marketplace per request, instead of relying on an API key or subscription. The outlet reported that Pearl Connect is not eligible for OLAS Activity Rewards, which Olas reserves at launch for autonomous agents meeting a threshold under its Proof of Active Agent staking mechanism. IT Brief Australia quoted David Minarsch, chief executive of Valory and a founding member of Olas, describing the product as a way to reduce tool switching and allow agents to pay one another directly. Those statements are attributed to the outlet and are not independently confirmed here.
為什麼這很重要
The reported launch connects AI coding agents with financial and blockchain actions, including prediction-market transactions and agent-to-agent payments. It also shows why financial permissions create a distinct safety problem: IT Brief Australia reported that cannot be fully eliminated and that fund loss remains possible.
The reported product matters because it places financial capability inside a workflow that is usually associated with writing and analysing code. A user could ask a coding agent to research a prediction market, obtain a service from another agent and prepare a transaction in the same session. That reduces friction, but it also brings financial authorization closer to the ambiguous instructions, external content and tool calls that coding agents must interpret. The significance is therefore not just crypto convenience; it is the combination of language-driven software work with actions that can move funds.
IT Brief Australia’s account highlights a potential change in how software services are paid for. Under the reported design, one agent could pay another for a narrowly defined service on a per-request basis. That could be useful where an agent needs a specialised result without a standing credential or subscription. At the same time, the source gives no evidence about pricing, settlement reliability, dispute handling, identity verification, accounting treatment or whether agent-to-agent payments work consistently in practice. Those unknowns limit what can be concluded about the model’s broader commercial importance.
The reported self-custody design also has practical implications. IT Brief Australia said private keys remain outside the AI session and that signing happens locally, which could reduce the consequences of an agent directly accessing key material. But keeping keys outside the session does not eliminate the risk that an agent prepares the wrong transaction or persuades a user to approve one. The outlet reported that attacks cannot be fully eliminated, that sensitive information on the user’s computer could be exposed, and that unintended actions or fund loss remain possible. No independent security review, incident history or loss estimate is provided.
The product’s first use case also connects AI agents to prediction markets, where automated research and transaction preparation can affect decisions involving money. IT Brief Australia cited Olas’s reference to Polymarket monthly trading volume of $4.3 billion, but the source does not independently verify that figure or establish how much activity Pearl Connect could generate. Nor does the report show that the product improves market research, prediction accuracy or trading outcomes. Its demonstrated public significance, based on the source alone, is the integration of agent assistance, blockchain execution and user-controlled signing—not proof of financial performance.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
接下來看什麼
The key questions are whether Pearl Connect remains limited to user-directed actions, how reliably its local signing and recovery controls work, and whether users adopt it beyond prediction markets. IT Brief Australia’s report is not independently confirmed here, and the source provides no external testing, transaction audit or evidence of live user performance.
The first issue to watch is whether Pearl Connect preserves the boundary between preparing an action and executing it. IT Brief Australia reported that Pearl Connect is intended for user-directed sessions and does not trade on its own, unlike Polystrat. Future reporting should establish exactly which steps require explicit user approval, whether transaction details are displayed in a trustworthy form, and how the system handles ambiguous instructions, failed transactions and retries. The source says an agent can detect a failed transaction and retry or notify the user, but it does not describe the approval interface or safeguards against repeated or misdirected attempts.
Security testing will be especially important. IT Brief Australia reported that remains a non-zero risk and advised users running other Pearl agents in auto-run mode to disable that setting while Pearl Connect is active. That warning indicates that the surrounding agent environment, not only the wallet code, affects the risk profile. Independent reviewers would need to test malicious instructions in market data, marketplace responses and connected smart contracts, as well as attempts to expose computer data or induce unintended transfers. The source provides no results from such testing and no details of an external audit.
Availability and scope are other open questions. The report says the beta supports Polygon and Gnosis, with one chain per Connect agent, and that users may create additional agents for other supported chains. It does not state whether the beta is open to everyone, what geographic or account restrictions apply, what fees users pay, or how recovery works after a compromised machine or lost credentials. It also does not establish whether support for DeFi protocols is broad, curated or limited to specific contracts. These details will determine whether the product is a narrow experiment or a practical tool for regular users.
Finally, observers should distinguish adoption from promotional claims. IT Brief Australia reported that Olas pointed to prediction-market volume as evidence of market scale, but it supplied no Pearl Connect usage figures, transaction counts, independent user feedback or evidence of successful agent-to-agent payments. Follow-up coverage should seek those measurements, along with any reported incidents, changes to the beta’s permissions and evidence that users understand what they are authorizing. Based solely on the supplied report, Pearl Connect is a concrete beta launch with a consequential design direction, but its reliability, security and real-world impact remain unverified.