prom-evidence-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| PROMETHEUS_URL | Yes | Base URL of the Prometheus server | |
| PROM_MAX_SERIES | No | Maximum series count before cardinality warning | 200 |
| PROM_TIMEOUT_MS | No | Per-request timeout in milliseconds | 15000 |
| PROM_MIN_RANGE_SAMPLES | No | Minimum samples per series to avoid reliability warning | 5 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| instant_queryA | Run an instant PromQL query. The response includes an evidence block (exact query, execution time, series and sample counts, and warnings about empty or high cardinality results) that MUST be cited when making claims from this data. |
| range_queryA | Run a PromQL range query over a time window. The response includes an evidence block with sample coverage; if samples per series are low the evidence will say trends are unreliable, and that warning must be surfaced to the user. |
| list_metricsA | List metric names known to this Prometheus, optionally filtered by a substring. Use this to verify a metric actually exists before concluding anything from an empty query result. |
| target_healthA | Summarize scrape target health (up/down counts per job). Check this first during an incident: query results are only as trustworthy as the scrapes behind them, and a down target silently turns into missing data everywhere else. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 4 tools
Each tool targets a distinct operation: instant queries, range queries, metric listing, and target health. There is no overlap in purpose, and the descriptions reinforce when each should be used.
Names are lowercase snake_case and readable, with instant_query and range_query following a clear pattern. list_metrics and target_health deviate slightly from the adjective_noun style but remain predictable and unambiguous.
Four tools is well-scoped for a focused Prometheus evidence server, covering the essential query and health-check operations without unnecessary bloat.
The core workflow of verifying target health, confirming metric existence, and running instant/range queries is covered. Minor gaps exist such as label value enumeration or metric metadata, but agents can work around these with the available tools.