jev-harness
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| JEV_MODEL | No | Model override for the provider | |
| CMD_API_KEY | No | API key for Command Code provider | |
| VERCEL_API_KEY | No | Alternative API key for Vercel AI Gateway provider | |
| OPENCODE_API_KEY | No | API key for OpenCode Zen provider | |
| TYPESAFE_API_KEY | No | API key for TypeSafe AI provider | |
| AI_GATEWAY_API_KEY | No | API key for Vercel AI Gateway provider | |
| OPENROUTER_API_KEY | No | API key for OpenRouter provider (alpha access) | |
| COMMAND_CODE_API_KEY | No | Alternative API key for Command Code provider | |
| VERCEL_AI_GATEWAY_API_KEY | No | Another alternative API key for Vercel AI Gateway provider |
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_triage_test_failureA | Triages test traceback, compile error, or runtime failure using Jev System One (70-300ms, zero-generation). Returns root cause category, skip_llm flag (true if resolvable deterministically without frontier LLM), and immediate action recommendation. |
| jev_abort_checkA | Guards against doom loops, dead-ends, circular retries, and destructive refactors. Evaluates proposed plan against recent attempt history before burning tokens. |
| jev_route_taskA | Routes programming task to the minimal sufficient model tier (deterministic script, lightweight fast flash model, or heavy frontier reasoning model) to optimize cost and latency. |
| jev_verify_completionA | Calibrates step completion against acceptance criteria using typed rubric scoring. Checks if evidence is sufficient to declare done without launching expensive extra review loops. |
| jev_modulate_reasoning_effortA | Dynamically modulates reasoning effort (low, medium, high, etc.) and stability lease steps for the immediate generation step. Maps exact parameters for OpenAI (GPT-6 Astra/o3), DeepSeek (V4.1-Flash/R1), Qwen (3.8 Max), Anthropic (Claude Fable 5.1), and Gemini (3.8 Thinking). Eliminates reasoning token waste and cuts multi-minute delays on mechanical tool calls. |
| jev_should_nudge_continuationA | Evaluates whether an autonomous agent paused prematurely with unfinished work or unverified changes (Workflow phases: research, ask, plan, execute, verify, complete + CommandCode Jev Nudge protocol). Vetoes nudges when waiting on user permission/input or when the previous nudge produced no progress. |
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 6 tools
Each tool addresses a distinct decision point in the agent lifecycle: failure triage, abort protection, task routing, completion verification, reasoning effort modulation, and continuation nudging. There is no meaningful overlap; even the related pairs (abort vs. nudge, route vs. modulate) focus on different moments and inputs.
All tools share the 'jev_' prefix and use snake_case with a verb-like leading word, e.g., triage_test_failure, route_task, verify_completion. The only slight deviation is 'jev_should_nudge_continuation', which uses a modal verb phrase rather than a direct imperative, but it remains clear and predictable.
Six tools is a tight, purposeful set for a decision-harness server. Each tool covers a distinct optimization decision without redundancy or bloat, making the count well-scoped for the domain.
The harness covers the main meta-decision points: triage failures, prevent wasted work, route to cost-effective models, verify completion, modulate effort, and decide on nudges. A minor gap might be a tool to record outcomes or update state, but the core lifecycle appears well covered.