General Decision MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| JEV_MODEL | No | Native model name for the compatible provider. | |
| JEV_API_KEY | No | Gateway downstream key for the compatible provider. | |
| JEV_PROVIDER | No | Decision provider to use. Supported values are compatible and typesafe; cloudflare is not implemented. | |
| JEV_MCP_MODEL | No | Legacy alias for JEV_MODEL. Conflicting values fail startup. | |
| JEV_API_BASE_URL | No | Full native System One endpoint for the compatible provider. The endpoint is sent exactly as configured. | |
| JEV_API_KEY_FILE | No | File containing JEV_API_KEY. Mutually exclusive with JEV_API_KEY. | |
| TYPESAFE_API_KEY | No | API key for the typesafe provider. | |
| JEV_MCP_TRANSPORT | No | MCP transport setting. The --http CLI flag overrides this value. | |
| DECISION_HTTP_HOST | No | HTTP bind host for HTTP mode. | |
| DECISION_HTTP_PORT | No | HTTP bind port for HTTP mode. | |
| DECISION_MCP_TOKEN | No | Independent random Bearer token of at least 32 characters, required for HTTP mode. | |
| TYPESAFE_API_KEY_FILE | No | File containing TYPESAFE_API_KEY. Mutually exclusive with TYPESAFE_API_KEY. | |
| DECISION_MAX_IN_FLIGHT | No | Maximum number of in-flight calls. | |
| DECISION_MCP_TOKEN_FILE | No | File containing DECISION_MCP_TOKEN. Mutually exclusive with DECISION_MCP_TOKEN. | |
| DECISION_HTTP_ALLOWED_HOSTS | No | Comma-separated exact client-facing authorities allowed for non-loopback or container HTTP binds. | |
| DECISION_HTTP_ALLOWED_ORIGINS | No | Comma-separated exact origins allowed for requests with an Origin header. |
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
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| jev_coding_loopA | Suggest a bounded coding-loop next step from supplied observations and execution facts. Advisory only; never grants authorization or executes actions. |
| jev_reviewA | Assess supplied patch text against request and tests; all rubric dimensions must independently clear fixed floors. Advisory only; never grants authorization or executes actions. |
| jev_gateB | Combine independent patch-review dimensions and completion-claim checks; the weakest judgment governs and never authorizes deployment. Advisory only; never grants authorization or executes actions. |
| jev_riskB | Assess bounded infrastructure changes with deterministic human-review floors for security and unknown facts. Advisory only; never grants authorization or executes actions. |
| jev_verifyA | Check each supplied claim and its exact subject independently against supplied evidence. Advisory only; never grants authorization or executes actions. |
| jev_screenA | Estimate injection, relevance and substance signals in supplied text. Screening does not sanitize content or make it trusted. Advisory only; never grants authorization or executes actions. |
| jev_classifyB | Classify supplied items into bounded caller-defined classes, with an unavoidable investigate option. Advisory only; never grants authorization or executes actions. |
| jev_rankA | Rank bounded candidates by native Choice probabilities, preserving dependencies and surfacing any probability-order violations. Advisory only; never grants authorization or executes actions. |
| jev_decideA | Choose among 2–6 bounded alternatives using supplied evidence and priorities, with fixed escape options and independent requirement checks. Advisory only; never grants authorization or executes actions. |
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 9 tools
Each tool covers a distinct decision-support function (screen, classify, rank, verify, review, gate, risk, decide, coding loop), and the descriptions clarify the boundaries fairly well. The main friction is the review/gate pair and the verify/decide/rank cluster, which sit close together conceptually and require reading descriptions to separate.
All names share a consistent jev_ prefix with short descriptive suffixes, which is highly predictable. Minor deviation: jev_coding_loop is a compound noun phrase while the rest are single verbs, but it remains readable and unlikely to confuse.
Nine tools is well within a comfortable scope for a decision-support suite, and each maps to a different stage of the pipeline rather than duplicating another. No tool appears redundant or trivially filler.
The surface covers the full advisory pipeline from screening and classification through ranking, verification, review, gating and final decision, which is solid coverage for the stated purpose. Minor gaps exist around recording/auditing decisions or retrieving prior judgments, but agents can work around these by passing context in.