Jev MCP on Viorant

Viorant exposes TypeSafe's Jev decision model as a remote MCP connector, so MCP clients can give it context and a question and get back a choice, a score or a yes/no with a probability to act on. The connector is at https://jev.viorant.io/mcp and authenticates with OAuth.

Six: jev_noul, jev_choice, jev_score, jev_verify, jev_screen and jev_find. The first three are the primitives that return typed answers with probabilities, and the other three compose them for verification and screening tasks.

For Claude Code, run claude mcp add --transport http jev https://jev.viorant.io/mcp and then run /mcp and choose Authenticate for jev. Cursor and VS Code take the same URL in their MCP configuration, or you can add it as a connector in the Viorant Hub.

The page lists routing to cheaper models, verified cascades that escalate only when it matters, triage scoring for support tickets, runtime gates that screen fetched pages or tool results for prompt injection, and claim checks against evidence.

The Free Plan includes 2,500 Vio Credits, which Viorant quotes as up to 25M JEV input tokens. The page is explicit that this is a theoretical equivalent subject to rounding and other credit usage, so treat it as a marketing figure and follow the terms.

Yes. The page documents the Claude Code command, a Cursor mcp.json block, a VS Code servers configuration and adding the connector URL in the Viorant Hub. The vio CLI path — vio docs agents, vio deploy and vio connect — is one option, not a requirement.

The captured excerpt is truncated inside the jev_verify description, so the full tool reference, pricing details and rate limits are not recorded in this entry. Check the live Viorant page before building against it.