MCP server for Platoona Connect - Access 10,000+ SaaS integrations from AI assistants like Claude Code, Cursor, and more
Platoona MCP has well-structured tool definitions with clear naming (verb_noun pattern) and comprehensive parameter schemas using Zod. All 6 tools have descriptions and input schemas. However, output schemas are partially documented (search-tools and list-integrations have outputSchema in code, but others lack explicit return type documentation). Parameter descriptions are present but some lack actionable constraints (e.g., execute-tool's 'parameters' object is generic). Error handling returns structured responses but lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in metadata.
Connect to an integration using OAuth or API key authentication. For OAuth integrations, returns an authorization URL. For API key integrations, creates the connection directly.
Disconnect from an integration and revoke access.
Execute an integration action with the specified parameters. The action must be from a connected integration.
List all active integration connections. Use this to see which integrations are already connected.
List available integrations/apps that can be connected.
Search for available integration tools/actions using semantic search (RAG). Use this to discover what tools are available for a specific task or integration.
execute-tool's 'parameters' field is a generic object with no schema constraints. LLMs cannot infer what parameters an action expects without calling search-tools first. This breaks composability and forces discovery overhead.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications in metadata (READ_ONLY, WRITE, REVERSIBLE). Agents cannot infer safety properties without explicit hints.
list-connections has empty input schema ({}). Description is minimal (18 chars). No guidance on what the response structure contains or when to call it.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2025-06-18+ | v2 |
Error responses return isError=true with text content but lack recovery guidance. E.g., 'Error searching tools: Unknown error' tells the agent nothing about retry eligibility or next steps.
connect-app accepts optional 'scopes' and 'redirectUrl' but does not document OAuth vs API key branching logic. LLMs cannot predict which parameters apply to which integration type.