Solanaprox

PassAudited by ClawScan on May 10, 2026.

Overview

SolanaProx is a disclosed paid AI proxy that does not ask for private keys, but it sends prompts and a wallet address to third parties and can spend pre-funded balance when used.

Before installing, make sure you are comfortable with a third-party proxy seeing your prompts, responses, and public wallet address. Fund only a limited balance, request cost/balance confirmation for repeated calls, and do not run the optional npx MCP server or AIProx registration command unless you have reviewed and trust those separate components.

Findings (5)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

When you ask the agent to use SolanaProx, successful calls may reduce your deposited USDC/SOL balance.

Why it was flagged

The skill's core tool use triggers paid AI requests, so use can spend a pre-funded balance. This is disclosed and purpose-aligned, with balance-check guidance.

Skill content
- Pay-per-request pricing in USDC or SOL
Recommendation

Fund only the amount you are comfortable spending and ask the agent to confirm model, estimated cost, and balance before multi-step or repeated use.

What this means

Your public wallet address is shared with SolanaProx and linked to your requests and balance, but the skill does not request your seed phrase or private key.

Why it was flagged

The wallet address is used as the authentication identity for the service. The artifact explicitly says private keys are not accessed.

Skill content
The wallet address is sent as an HTTP header — no private keys or seed phrases are required or transmitted.
Recommendation

Use a wallet address you are comfortable associating with these AI requests, and never provide private keys or seed phrases.

What this means

Any sensitive information included in prompts may be visible to the proxy service and upstream AI providers.

Why it was flagged

The artifact clearly discloses that user prompts and outputs are sent through SolanaProx and then to upstream model providers.

Skill content
All prompts and responses pass through this proxy to upstream model providers (Anthropic, OpenAI).
Recommendation

Avoid sending confidential, regulated, or wallet-sensitive information unless you trust the proxy and provider data handling.

What this means

If you run the optional MCP server command, you are trusting code that was not part of this scan.

Why it was flagged

The document includes an optional command that would fetch and run an npm package not included in the reviewed artifacts.

Skill content
npx solanaprox-mcp
Recommendation

Inspect the npm package, pin a trusted version, and run it only if you intentionally want the MCP server.

What this means

A user could overlook that the optional registry workflow contacts a separate site and may publish agent registration details.

Why it was flagged

The security manifest says only solanaprox.com is called, but the documentation also provides an optional registration command to aiprox.dev.

Skill content
External endpoints called: https://solanaprox.com/ (only) ... curl -X POST https://aiprox.dev/api/agents/register
Recommendation

Treat AIProx registration as a separate, optional action and confirm what information will be sent before using it.