fix: return graceful response when AGENT_WALLET_SERVER is not configured

[Rohiteswar]

Problem

Users setting up this MCP server for the first time are confused by AGENT_WALLET_SERVER including myself. The README described it as a URL but wasn't clear it's a separate server you run yourself, not an API key or an Alchemy-provided value.

When AGENT_WALLET_SERVER was unset, calling sendTransaction or swap would silently crash with a cryptic network error (fetch failed or TypeError: undefined/transactions/send) — giving users no indication of what went wrong or how to fix it.

Changes

• sendTransaction and swap now return a graceful { status,message } response when AGENT_WALLET_SERVER is not set, instead of network call crashing. The MCP server keeps running and Claude can relay the message clearly to the user or the dev.
• Clarified the README to explain that AGENT_WALLET_SERVER is the base URL of a separately running HTTP server (e.g.http://localhost:3000),not an API key.
• Added before-and-after screenshots showcasing the improved error handling, replacing unclear network failures with structured { status, message } response.

Screenshots

• Before: When AGENT_WALLET_SERVER was not configured, the MCP server did not handle the request gracefully, resulting in no meaningful response being returned to Claude, which made the issue unclear to the user.

• After: With the updated error handling, the MCP server actively returns a structured { status, message } response. Claude can now surface this response through the tooling interface, providing clear and actionable feedback to the user or developer.

Why graceful response over throw?

A hard throw would surface as a red error in the agent. Returning a structured response lets Claude read it and explain the situation naturally to the user — better UX with no server interruption.

Testing

pnpm build passes ✓