Agentic SRE Platform with LangGraph orchestration, MCP integration, and A2A agent routing for Kubernetes, metrics, and incident management
Server has 16 tools with basic naming and descriptions, but significant quality gaps. Most tools lack comprehensive parameter descriptions and output schemas. Parameter formats are documented in descriptions (e.g., 'namespace:pod_name') rather than enforced via schema constraints. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Error handling and recovery guidance absent. Descriptions average ~100 chars but lack LLM-optimized context about when/why to use each tool.
Fetch and analyze CPU, memory, error rate, and latency metrics from Datadog and AWS CloudWatch.
Delegate the same task to ALL agents that have a given capability and return a list of results.
Close a ServiceNow incident with resolution notes.
Create a new ServiceNow incident from an SRE alert.
Submit input data to the best agent that supports a given capability.
Detect metric anomalies using statistical methods (z-score analysis).
These tools provide no guidance on what parameters they accept or require.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels declared (READ_ONLY, REVERSIBLE, WRITE). LLMs cannot infer safety properties from risk labels alone; annotations must be in the schema.
Parameter formats documented in descriptions (e.g., 'namespace:pod_name:tail_lines') rather than enforced via JSON Schema constraints. LLMs cannot parse free-form format strings reliably; use regex patterns, enums, or structured objects.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | <=2025-11-25 | v2 |
Generate human-readable operational insights from the last metrics analysis.
Retrieve the current status and details of a ServiceNow incident.
Get the status of a Kubernetes node.
Fetch the most recent log lines from a Kubernetes pod container.
Retrieve the current status of a Kubernetes pod.
Return current status and output/error for a task.
Restart a Kubernetes pod by deleting it (the controller recreates it).
Determine which agent should handle an alert and delegate to it based on alert type and capability matching.
Scale a Kubernetes Deployment to a specified number of replicas.
Update an existing ServiceNow incident with work notes or status changes.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure. E.g., does get_pod_status return pod_id, namespace, status fields?
No error handling or recovery guidance. Tools lack descriptions of failure modes, retryability, or next steps. E.g., 'pod not found' should suggest search_pods() or list available pods.