MCP server for trading stocks through Zerodha exchange
Critical gaps across all dimensions. Tool 'add' lacks any description (0/100). Three trading tools ('buy-a-stock', 'sell-a-stock', 'sell-portfolio') have minimal descriptions (10-20 chars, well below 34-char baseline). ALL parameters lack descriptions entirely, none explain what 'stock', 'qty', or other inputs mean. Input schemas use Zod but provide no parameter-level documentation. No output schemas documented. Error handling is absent (try/catch blocks swallow errors silently). No validation of inputs (e.g., qty could be negative). Tool naming uses hyphens inconsistently ('buy-a-stock' vs 'sell-portfolio'). The 'sell-portfolio' description ('shows all the stocks owned') contradicts its name and risk classification (READ_ONLY is correct, but name suggests selling). No security controls: API keys hardcoded in trade.ts as string literals, not injected. No audit logging. Composition is poor: 'add' tool is unrelated to trading domain.
buy the stock through zerodha exchange
sells the stock on the zerodha exchange
shows all the stocks owned
All parameters lack descriptions. 'stock' and 'qty' in buy/sell tools have null descriptions; LLM cannot infer what values are valid or expected format.
API credentials hardcoded as string literals in trade.ts (apiKey, apiSecret, requestToken, accessToken). Secrets must never appear in source code or tool parameters; use environment variables or vault injection.
Tool 'add' has no description. Tool 'sell-portfolio' description is 24 chars ('shows all the stocks owned'), below 34-char baseline and does not explain when/why to use it.
No output schemas documented. Tools return unstructured text responses (e.g., 'Stock has been bought'). LLM cannot parse structured data or chain results to downstream tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | 2026-07-28+ | v2 |
Error handling is silent. placeOrder() and getPositions() catch errors but only log to console; no error messages returned to LLM. Agent cannot recover or retry intelligently.