ZendIQ

What ZendIQ does

The ZendIQ MCP server gives agents a structured way to inspect Solana tokens and assess swaps before a wallet signs anything. Its public MCP surface includes three tools: zendiq_screen_token, zendiq_triage_swap, and zendiq_optimize_swap.

Token screening is free and does not require a wallet, payment, API key, or local repository clone. Paid analysis uses mainnet USDC through x402. The returned contract includes a headline score and level, individual signals, observed values, status information, factor contributions, coverage details, and an analysis ID that can be reused by optimization for 60 seconds.

How it works

The ZendIQ MCP server runs locally and communicates with an agent over newline-delimited JSON-RPC on standard input and output. Diagnostic messages go to stderr so stdout remains a valid MCP transport. The adapter sends screening, analysis, and optimization requests to ZendIQ’s hosted API; the scoring engine itself is not included in the repository and has no local operating mode.

Payment signing happens on the user’s machine using a Solana keypair. A local decision ledger applies a hard spending ceiling, while the x402 facilitator handles payment settlement and network fees. The hosted API is the default destination, and calls are real rather than simulated: analysis costs $0.01 and optimization costs $0.02 in mainnet USDC.

Setup and configuration

Use Node.js 22.5 or newer. Install dependencies with npm ci, then run npm run budget:init to create a local ledger. The initialization reads the payment network from GET /v1/agent and does not spend funds.

Paid operation requires a user-controlled Solana keypair at runtime/payer-mainnet.key.json, funded with mainnet USDC. The README states that SOL is not required because the facilitator pays the payment transaction’s network fee. The server can be started with npm run mcp, or directly through npx -y @zendiq/mcp@latest.

The MCP endpoint can be changed with ZENDIQ_AGENT_URL. The examples and demo runner use ZENDIQ_API_URL instead; setting one does not change the other. ZENDIQ_AGENT_NETWORK controls the MCP server’s network selection and accepts mainnet or devnet, defaulting to mainnet.

Tools and capabilities

• Screen a Solana token without payment or wallet setup.
• Triage a proposed swap using ZendIQ’s hosted risk analysis.
• Build and verify an optimized unsigned swap through /optimize.
• Expose signal-level results instead of only a combined verdict.
• Apply a local budget ceiling to paid calls.
• Retry a settlement-pending payment through the client’s redemption flow.

Limitations and notes

The ZendIQ MCP server does not contain the scoring model. Signal derivation, data sources, thresholds, and weighting remain behind the hosted /analyse endpoint. Every score therefore depends on network access to ZendIQ’s API.

Optimization is build-only: it returns an unsigned transaction and does not execute a trade. The taker is a public key, but the resulting transaction must be signed by the wallet that controls it. Execution, when used by the broader repository tooling, can involve Jupiter and a Solana RPC.

Initial scans usually take one to five seconds, with some requests taking longer. Paid responses may wait up to 90 seconds for settlement confirmation, so an MCP host may need a longer tool timeout. Autonomous watch loops continue paying per candidate until the local ledger reaches its ceiling; the budget should be configured deliberately before running them.