An AI assistant can explain an investment. Getting it to act on that information requires something else: access to assets, permission to transact and a way to pay the network fees.
That is the problem Pharos Network is tackling with its Agent Native upgrade, introduced at Korea Blockchain Week on September 30.
The rollout brings together an AI interface for users, new developer tools and changes to how blockchain transactions are paid for. It also sets out a roadmap stretching into 2028 for agents that can operate more directly within financial applications.
For Pharos, the starting point is RealFi: bringing real-world assets and financial activity onchain. The next step is making that activity easier for people—and the software acting on their behalf—to navigate.
Post from @pharos_networkView the original post on X↗
Ask About Your Assets, Then Prepare an Action
The most immediate user-facing change is the Pharos Port Agent Widget, starting with Harbor.
Users will be able to ask about their real-world asset positions, including yield, when a vault matures and when a redemption window opens. Instead of navigating multiple screens to find that information, they can state what they want to know or do.
According to Pharos, the agent interprets the request and prepares the corresponding action, while the user retains control.
That distinction matters. Asking when an investment becomes redeemable is different from authorizing a transaction. The proposed experience connects the question with a possible next step without presenting the assistant as having unrestricted access to a user’s money.
Someone Still Has to Pay the Transaction Fee
At the protocol level, the upgrade introduces Sponsored Transactions. The mechanism separates the account initiating a transaction from the account paying its gas fee. An application or approved sponsor can cover the network cost while preserving the identity of the original sender.
For users, that could remove a familiar obstacle: needing to acquire a network’s gas token before using an application. For developers building agent-driven services, it provides a way to fund activity without requiring every acting account to manage its own gas balance.
The transaction still has a cost. Sponsorship changes who pays it. Pharos places gas sponsorship within its testnet rollout, with broader production deployment on its published roadmap.
Post from @pharos_networkView the original post on X↗
Giving Developers the Tools to Connect Agents
The new Pharos AI portal brings together Pharos Chain MCP, Developer Docs MCP and the official SkillEngine.
These tools are intended to help developers connect AI systems with network capabilities, technical documentation and reusable skills. The aim is to make those resources available from a common entry point rather than requiring teams to assemble each connection separately.
A parallel collaboration, AnvitaMate × Pharos, packages the capabilities into a local AI environment from infrastructure provider Anvita Flow. Developers can work with the Pharos tools within their own workspace.
Pharos also says a $10 million builder incubator supports early teams working across RealFi and AI.
A Roadmap Beyond the Chat Interface
Pharos’ Pacific Ocean mainnet launched in April 2026. The Agent Native announcement describes a further expansion of that network, with capabilities arriving in stages.
Its published roadmap places proof-of-stake staking on mainnet and gas sponsorship on testnet in Q4 2026. Mainnet gas sponsorship and specifications for agent accounts and transaction interfaces follow in the first half of 2027, with further integration later that year. An integrated Agent Native stack is the longer-term goal for 2028 and beyond.
“A brand refresh is not as simple as a new visual identity. It is about telling the story of what Pharos has been building: how RealFi brings real-world value onchain, how Agent Native changes the way people interact with that value, and how Pharos provides the infrastructure underneath." - Wish Wu, co-founder and CEO of Pharos Network
"We are not adding an AI feature to the application layer. We are giving agents — a new class of economic actors — the accounts, authorization and payments they need to operate. Today, those three lines came together for the first time," Wish added.
For users, the practical test will be whether this makes an investment easier to understand and manage while keeping permissions clear. For developers, it is whether the tools and protocol changes make agent-driven applications easier to build and operate.
A conversational interface is the visible part. The harder work is making sure the action behind the conversation is authorized, funded and executed as intended.
Reporting by Lidia Yadlos
