Agent payment infrastructure MCP server. Give AI agents autonomous payment capability, including stablecoin transfers, currency swaps, agent wallets, fiat operations, and real-time currency data via the rate service on Hedera. Works with Claude, LangChain, CrewAI, and AutoGPT.
40 tools with significant quality gaps. Most tools have basic descriptions (10-50 chars) that lack actionable context for LLM selection. Parameter descriptions are sparse or missing entirely. No input schemas visible in source for tools 25 (create_offer) and others inferred from mock/server.ts. Critical financial tools (transfer_tokens, swap_currency, execute_approved_transaction) lack error recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite high-risk operations. Naming is verb-first but many tools are ambiguous (e.g., 'update_offer' vs 'delist_offer' vs 'delete_offer', unclear distinctions). Output schemas undocumented. Baseline: 549 production tools average 194-char descriptions; this server averages ~60 chars.
Cancel an active trade
Check the status of an approval request
Create a new offer for trading
Delete an offer permanently
Delist an offer (make it inactive)
Echoes the input message back to the caller
Engage with an offer to create a trade
create_offer (tool 25) has NO visible input schema in source code. Schema score must be 0. Tool definition inferred from mock/server.ts only, capping overall at 50.
Irreversible/WRITE tools (transfer_tokens, swap_currency, execute_approved_transaction, delete_offer) lack error recovery guidance. Descriptions do not explain what to do if the operation fails, how to verify success, or whether it is idempotent. LLMs cannot self-correct on failure.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite 40 tools with varying risk levels. LLMs cannot distinguish safe reads from destructive writes without explicit hints. Critical for financial operations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Estimate fees for a currency swap before executing
Execute a transaction that has been approved
Get configuration for an agent user
Get Hedera account balances for all tokens or a specific token
Get the status of a fiat deposit
Get escrow balances for trades
Get escrow transaction history
Get current buy/sell exchange rates for a currency
Get KYC verification status for a user
Get your API key limits and current usage
Get all offers created by the current user
Get your BananaCrystal profile including user ID, wallets, and account details
Get all trades for the current user
Get details of a specific offer
Returns basic server information
Get details of a specific trade
Get your transaction history with filtering options
Get the status of a fiat withdrawal
Initiate a fiat deposit via Brale. Requires verified KYC.
Initiate KYC verification for a user. Returns a Sumsub access token.
List all available Hedera tokens for the platform
List all available offers with optional filtering
List all supported currencies and tokens on BananaCrystal
List all trades with optional filtering
Health check tool that returns pong
Request an agent-to-agent transaction that requires approval
Step 1: Request an OTP code to authorize a token transfer. An email will be sent with the code.
Request a fiat withdrawal via Brale. Requires verified KYC.
Reset your sandbox balance to initial state (sandbox only)
Swap between two Hedera tokens. Exchange rate calculated automatically.
Step 2: Execute the token transfer. If OTP is enabled on your key (default), provide otpCode and transactionRef from request_transfer_otp. If OTP is disabled, these fields are not required.
Update agent approval settings (require_human_approval, require_otp, callback_url)
Update an existing offer
Deposit/withdrawal tools (initiate_deposit, request_withdrawal, tool 18-19) have minimal descriptions (35 chars) and sparse parameter documentation. Missing details on KYC requirements, supported currencies, and rail types. LLMs cannot determine when to use these vs transfer_tokens.
Output schemas undocumented across all 40 tools. No visible documentation of response structure, field types, or pagination. LLMs cannot plan downstream tool calls or extract required IDs (e.g., transactionId, offerId) for chaining.