DecisionGate
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| DECIDER_URL | No | URL of a running strands-decider server, required when using the decider-http scorer. | |
| GEMINI_API_KEY | No | API key for the Gemini scorer; required when using the gemini scorer. | |
| DECISIONGATE_SCORER | No | Scorer backend to use: baseline, fake, decider-http, decider-local, or gemini. Defaults to a keyless scorer. | |
| DECISIONGATE_FLAG_AT | No | Threshold for flagging a tool call. | |
| DECISIONGATE_ALLOW_AT | No | Threshold for allowing a tool call based on p_safe. | |
| DECISIONGATE_ON_ERROR | No | Behavior when the scorer errors; defaults to failing closed and blocking. | |
| DECISIONGATE_BLOCK_BELOW | No | Threshold below which a tool call is blocked based on p_safe. | |
| DECISIONGATE_INJECTION_BLOCK_AT | No | Threshold at or above which a tool call is blocked based on p_injection. |
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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| score_tool_callA | Score a PROPOSED tool call before running it. Returns action allow | ask | block,
p_safe, p_prod, p_injection, the allow/ask/block choice distribution, a 0-4 risk
score and reasons. |
| gate_infoB | Which scorer and thresholds this gate uses, and the questions it asks the model. |
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 2 tools
The two tools have clearly distinct purposes: score_tool_call evaluates a proposed tool invocation, while gate_info describes the scorer, thresholds, and evaluation questions. An agent should have no trouble selecting between them.
Both names use snake_case and are readable, but score_tool_call follows a verb_noun action pattern while gate_info is a noun phrase for metadata. This is a minor convention deviation rather than a confusing inconsistency.
Two tools is thin for a server, with only one core functional operation and one informational tool. For a narrowly scoped decision gate this is defensible, but it sits at the lower edge of what feels complete.
The core gate workflow is covered: score a proposed call and inspect the gate's scorer and thresholds. Minor gaps remain around batch scoring, audit/history, or policy inspection, but agents can work within the current surface.