Production-ready MCP server for Argo Workflows
Argo Workflows MCP server demonstrates good definition quality with clear verb-starting tool names, comprehensive parameter schemas, and detailed descriptions. All five tools have explicit input schemas with typed parameters and descriptions. However, output schemas are not documented in the visible source code, and several tools lack error handling guidance. The tool set is well-composed for Argo-specific workflows without redundancy. Most parameters include sensible defaults and constraints (enums, ranges), though some descriptions could be more action-oriented for LLM guidance.
Get detailed information about a specific workflow
Check the health and connectivity of the Argo server
List workflows with advanced filtering, pagination, and sorting options
Submit a new workflow from a template
Get logs from a workflow
Output schemas not documented in source. Tools return structured data (workflows, logs, etc.) but the expected response shape, field types, and required fields are not visible in the provided code. LLMs cannot plan downstream calls without knowing what fields to extract.
Error handling guidance missing. Tools have no documented error cases (e.g., 'workflow not found', 'invalid namespace', 'Argo server unreachable'). Error responses should guide the LLM on recovery: retry, adjust params, or fail. Without this, agents hit errors and stall.
submit_workflow description lacks state-change clarity. Description is 'Submit a new workflow from a template' but does not explicitly state this is a WRITE operation with irreversible consequences. Agents need to know this cannot be safely retried without side effects. The Risk field shows WRITE, but the description should reinforce this.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | A | 80 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 48 | - | v1 |
workflow_logs 'follow' parameter has no guidance on behavior. Description says 'Follow log stream' but does not explain if this streams indefinitely, returns a snapshot, or uses SSE. LLMs cannot predict the tool's blocking behavior without this detail.
list_workflows parameter relationships undocumented. Parameters like 'offset', 'continue', and pagination token semantics are not explained. Which takes precedence? What is the max page size? LLMs will guess, causing inefficient pagination.
submit_workflow 'parameters' param is additionalProperties: true with no constraints or examples. This is a free-form object, LLMs cannot know what keys to pass without documentation. Should document expected keys, types, and sample structure.