jev-mcp-server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@jev-mcp-serverClassify 'I want to wash my car, the wash is 50m away' as question, command, smalltalk, or info."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
jev-mcp
Minimal MCP server exposing TypeSafe/OpenRouter /api/alpha/decisions as
five judgment tools for coding agents.
Tools
Tool | Purpose | Returns |
| Does the evidence support the claim? | noul (0..1) + confidence |
| Which label fits the state? | choice + probabilities + confidence |
| Relevance + risk for fetched content | choice + score |
| Pick from rich options with criteria | choice + probabilities |
| Detect named fields in a state | per-field noul |
Related MCP server: jev-code
Install (run)
bun install
bun run buildRegister with an MCP client
OpenCode / Claude Code:
{
"mcpServers": {
"jev": {
"command": "node",
"args": ["/path/to/jev-mcp/dist/index.js"],
"env": {
"OPENROUTER_API_KEY": "sk-or-v1-..."
}
}
}
}Or use TYPESAFE_API_KEY (TypeSafe direct, cheaper) instead of OPENROUTER_API_KEY.
The MCP client passes tool inputs as JSON; the server returns JSON text content
plus isError: true when fail_on predicates match.
Wire example
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "jev_classify",
"arguments": {
"state": "I want to wash my car, the wash is 50m away",
"choices": ["question", "command", "smalltalk", "info"]
}
}
}Response:
{
"jsonrpc": "2.0",
"id": 1,
"result": {
"content": [
{
"type": "text",
"text": "{\"type\":\"choice\",\"choice\":\"info\",\"probabilities\":{\"info\":0.96,...},\"confidence\":0.94}"
}
]
}
}Tool input contracts
See src/tools.ts for the JSON Schema for each tool. Highlights:
jev_verify—{ claim, evidence, instructions?, fail_on?: "true"|"false", timeout_ms? }jev_classify—{ state, choices: string[], criteria?, instructions?, fail_on?, timeout_ms? }jev_screen—{ text, relevance_labels?, fail_on?: { relevance?, min_risk? }, timeout_ms? }jev_decide—{ state, options: Record<key, description>, instructions?, fail_on?, timeout_ms? }jev_extract—{ state, fields: string[], timeout_ms? }
fail_on predicates return isError: true (rather than silently passing) when
matched, so agents see the warning and decide whether to proceed.
Design notes
Single-binary, zero runtime deps. Pure Bun stdlib +
node:fs/promises.Provider abstraction at the wire level (TypeSafe direct or OpenRouter via the
/api/alpha/decisionsendpoint). Routing picks TYPESAFE if set, otherwise OR.No telemetry, no state across calls. Each tool call is independent.
Original implementation. Design intent informed only by the public MCP protocol and the public TypeSafe/OpenRouter API contracts.
License
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
Pre-execution governance for AI agents. Deterministic PASS/FAIL/REVIEW verdicts, replayable proof.
Deterministic runtime safety for AI agents: scan PII, gate tool actions, verify LLM output.
- SnipgetOAuthai.snipget
300+ deterministic data utilities for AI agents: validate, normalize, parse, match, redact.
See, price, and control every tool call your AI agents make: policy checks, cost, and audit tools.
Related MCP Servers
- AlicenseAqualityCmaintenanceProvides agents with fast, typed, calibrated decision tools for classification, scoring, yes/no checks, and gating risky tool calls.5981 npmMIT
- AlicenseNot gradedqualityCmaintenanceProvides coding agents with typed classification, yes/no checks, scoring, ranking, and question-answering tools that return calibrated probabilities for fast, reliable decisions.9MIT
- AlicenseNot gradedqualityCmaintenanceEnables coding agents to send natural language and JSON application state to a judge tool that returns typed judgments (yes/no probability, choice, or score) with model and usage metadata, and fails gracefully without blocking workflows.1Apache 2.0
- AlicenseAqualityCmaintenanceEnables agents to call typed code-review and content-moderation decision tools, returning structured verdicts, probabilities, and confidence-gated actions.2MIT