Pranav Maheshwari

Pranav Maheshwari works on the practical connections between infrastructure and the applications that use it: blockchain data, programmable wallets, and agentic commerce. Identified by Edge & Node as Director of Integrations, he co-presented Ampersend with Rodrigo Coelho as a way to connect AI agents to paid services and transaction controls. The problem is concrete: an agent may know what a task requires yet lack a way to purchase the information or service needed to finish it.

Connecting blockchain infrastructure to developers

Maheshwari’s developer-facing work predates that focus on agents. Scaling Ethereum’s 2021 program identified him with Polygon, teaching alongside Arjun Kalsy. By October that year, The Graph’s coverage of a Dubai developer event identified him as a solutions engineer at Edge & Node. The Graph indexes blockchain records so applications can retrieve and use them through queries. His work connected that infrastructure to developers building applications on it.

He also argued that digital ownership depends on more than recording an asset on a blockchain. An NFT application can still have a single point of failure if its surrounding content and services depend on a central server. Decentralized storage and data access help the application preserve the properties its ownership records promise. This concern links his infrastructure work to the user’s experience of an asset: ownership is less useful when the application needed to access it can disappear.

In 2023, ETHSeoul’s lineup listed Maheshwari as a Senior Solutions Architect at Edge & Node. The Graph’s builders newsletter also identified him with developer relations and as the host of Web3KiDuniya, extending his technical work into education and community participation.

Programmable wallets need readable account histories

His June 2023 explanation of account abstraction examined how programmable wallets could make blockchain applications easier to use. Session keys, recurring payments, recovery mechanisms, and spending limits give accounts behavior beyond signing individual transactions. Those features also require understandable records. Indexing transfers, withdrawals, sponsored operations, and recovery events turns smart-contract activity into an account history a user can inspect. His examples included tracking Biconomy paymaster deposits, withdrawals, and sponsored operations, connecting wallet convenience to the data infrastructure underneath it.

Ampersend brings payments into agent workflows

His work on Ampersend brings tool access and payments into an agent’s workflow. A supplied event listing identified him as its AI Engineering and Automations Lead. In Agent Spending Without Controls, presented with Coelho, the pair demonstrated how an aggregation layer could give agents access to paid Model Context Protocol (MCP) tools through an installed skill file. Rather than requiring a user to register and provide billing details separately for every service, the system handles payments as the agent uses those tools.

• Paid information access: The presentation compared the same information request in two agent environments, one with the Ampersend skill file and one without it. The paid-tool environment returned more specific information, and the presenters showed an authorized transaction for the endpoint used. The example illustrates why payment access can affect an agent’s capability: a model’s ability to reason does not give it access to every useful source. Their expectation that more valuable MCP services will charge for access is a position about the market’s development, rather than an established outcome.
• Shopping and budget limits: A second demonstration carried product discovery, checkout, and an order receipt through a terminal. The gift-shopping request specified a budget below $10, but the spoken walkthrough later reported an $11 charge. The example therefore supports the possibility of completing commerce within an agent workflow while leaving reliable budget enforcement unresolved. The presenters also described how remembered preferences and reusable payment details could reduce repeated forms and account setup.
• Wallet screening: Their wallet-screening simulation addressed the seller’s decision to accept a payment. With screening disabled, two agents could purchase access to a scraping service. Enabling a TRM-based integration caused the flagged wallet’s transactions to be rejected while the other agent continued. This demonstrated a specific control in a simulation: screening a wallet can prevent a payment from being authorized. It did not establish comprehensive compliance or protection against every form of agent error.

Maheshwari’s work connects application usability to the infrastructure that makes it possible. Indexed events make programmable accounts understandable; payment access lets agents acquire useful services; transaction controls determine which purchases a system will accept. His collaboration on Ampersend develops those connections through concrete workflows, including their unresolved limits.