Loading ToolBench
MCP server that provides tools for querying and managing Chronosphere observability platform, including Prometheus metrics, logs, traces, monitors, and configuration management
Chronosphere MCP provides 7 well-named Prometheus query tools with clear verb-noun naming (query_*, list_*, render_*). All tools have descriptions (avg 150 chars) and documented input schemas with types. Parameters include descriptions and sensible defaults (limit=100, offset=0, step_seconds=60). However, output schemas are not explicitly documented in the source, and error handling guidance is absent. No tool annotations (readOnlyHint, destructiveHint) despite all being read-only. Parameter descriptions lack format constraints (e.g., RFC3339 format for time params is mentioned but not formalized). No pagination guidance in descriptions despite limit/offset support.
Returns the list of label names (keys) available on metrics that match the given selectors. Use this tool when you need to discover what labels are available on specific metrics or services. This returns only the label keys (e.g., "method", "status", "service"). To get the values for a specific label, use list_prometheus_label_values.
Returns the list of values for a specific label name, optionally filtered by selectors. Use this tool when you know the label name and want to discover what values it has across your metrics. Supports pagination with limit/offset.
Returns the complete time series (full label sets with all key-value pairs) that match the given selectors. Each result shows the exact combination of labels for an active time series. Use this tool only when you need to see the actual label combinations that exist. Supports pagination with limit/offset to manage large result sets.
Returns metadata for the given metric in the form of "type", "help", and "unit". If the metric does not exist or there is no metadata for this metric, return an empty object.
Output schemas not documented. Tool descriptions state what is returned (e.g., 'time series data points as JSON', 'metadata for the given metric') but do not specify the structure of response objects, field names, or types. LLMs cannot plan downstream operations without knowing response shape.
No tool annotations despite all tools being read-only. Missing readOnlyHint annotation on all 7 tools. This prevents clients from optimizing caching, retry logic, and safety checks for read-only operations.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | <=2025-11-25 | v2 |
Evaluates a Prometheus instant query at a single point in time
Executes a Prometheus PromQL query over a specified time range and returns time series data points as JSON. Supports standard PromQL syntax plus Chronosphere custom functions: cardinality_estimate(vector), head_avg/head_max/head_min/head_sum(query, n), tail_avg/tail_max/tail_min/tail_sum(query, n), sum_per_second(range_vector). Returns raw time series data - use query_prometheus_instant for single point-in-time values.
Evaluates a Prometheus expression query over a range of time and renders it as a PNG image.
Time parameter format constraints not formalized. Descriptions mention 'RFC3339 format or timestamp in seconds' but do not use JSON Schema format/pattern fields. LLMs cannot validate format compliance and may pass invalid timestamps.
No error handling guidance. Tool descriptions do not explain what happens on invalid queries, missing metrics, or API failures. LLMs have no recovery path if a query fails.
Pagination not explained in tool descriptions. Tools accept limit/offset but descriptions do not state whether results are capped, how to iterate, or what happens when limit=0. Agents may not know pagination is available.