Jev directory
CommunityJev Agents & MCP
Use Jev
An MCP server and CLI that put Jev in front of any agent with escalation, an action gate, and four interchangeable backends — TypeSafe, OpenRouter, Vercel AI Gateway, or an offline mock — plus a zero-token Claude Code hook.
- Reading time
- 3 min
- Published by
- Community
- Author
- silkyland
- Last verified
- Oct 2026
- Languages
- JavaScript
- Platforms
- MCPCLIOpenRouterVercel AI Gateway
- Capabilities
- Requires API key
Highlights
- Tools: jev_ask, jev_noul, jev_choice, jev_score, jev_gate (allow/deny/ask), plus free local jev_route, jev_models, jev_status, and jev_howto.
- Backends: TypeSafe, OpenRouter (/api/alpha/decisions, alpha path), Vercel AI Gateway, and a deterministic offline mock; auto picks the first credential.
- Every question gets a verdict; oversized, malformed, and unreachable are no-judgment reasons and a gate fails open on all three.
- The Claude Code PreToolUse hook only ever tightens permissions and clips tool input at 4,000 characters by default.
- The README calls it a rate win, not a token win: data pasted into jev_ask travels through your context twice.
Watch out
The listing shows no licence. Only the TypeSafe backend is verified against live responses — the OpenRouter and Vercel paths are implemented from documentation, and the OpenRouter decisions path is alpha and may move. Headless runs need a permission rule because no permission mode auto-allows an MCP tool.
Questions about this entry
Short answers drawn from the linked project or page.
An MCP server and CLI that expose TypeSafe's System One API to MCP-capable agents, with escalation, an action gate, and four backends. Jev takes a state plus typed questions and returns typed answers with calibrated probabilities; it is not a chat or code model.
Run npx -y @silkyland/use-jev install: it detects the agents present, writes an MCP entry for each, links the bundled routing skill, and backs up every file it touches, never writing a credential. Add --dry-run to preview, --agent to narrow, or --no-skills for tools only; uninstall reverses it.
typesafe (api.typesafe.ai/v1/systemone), openrouter (an alpha decisions path), vercel (AI Gateway evaluation model), and mock for offline deterministic runs. JEV_BACKEND picks one; unset means auto, taking the first credential in the order typesafe, openrouter, vercel.
It fails open, silently with a line on stderr, because a judgment sidecar being down must never block the agent. The gate only ever tightens: deny becomes deny with a reason, unsure becomes ask, and allow produces no output, so your normal permission flow decides. It is not enabled by install.
oversized (state over about 30k tokens), malformed (an unreadable answer), and unreachable (provider failure) are exported as NO_JUDGMENT_REASONS, and a gate must fail open on all three. Other reasons include open_ended questions and answers below the confidence threshold.
use-jev judge --questions-file q.json judges data already in a file or piped on stdin, so items never enter an agent's context. It exits 0 when every verdict stands and 3 when any escalated, so a script can branch on the result.
confidenceFrom says where the number came from: reported is Jev's own head (choice and score only, threshold 0.5), and estimated is the top-minus-runner-up margin (threshold 0.4). Both are jev-use's calibration and can be retuned with --threshold, JEV_CONFIDENCE_THRESHOLD, or confidenceThreshold in config.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
Viorant hosts TypeSafe's Jev as a remote MCP connector any MCP client can call: six tools cover typed decisions and compositions — jev_noul, jev_choice, jev_score, jev_verify, jev_screen and jev_find — with 2,500 Vio Credits in the Free Plan.
Jev Agents & MCP#community#catalog-import#yibie
Community
An MCP proxy and hooks adapter that screens an agent's tool calls before they run and the tool results before the agent reads them, with thresholds in a policy file.
Jev Agents & MCP#community#mcp#security
Official MCP server for JevGuard and TypeSafe AI: zero-dependency guardrails over JSON-RPC 2.0 on stdio, gating shell commands and code patches, caching decisions in local SQLite and pruning state.
Jev Agents & MCP#community#catalog-import#cobanov
Community
seb4ez/jevguard-mcpPython
seb4ez/jevguard-mcpPythonMIT-licensed Python tool router for coding agents that keeps obvious calls on a fast path, asks Jev (default typesafe/jev-1.13 via OpenRouter) for genuinely ambiguous choices, and can run bounded MCTS when multi-step consequences matter. Ships a CLI, MCP server and hook integrations.
Jev Tools & CLIs#community#routing#mcp
A guard for coding agents that puts Jev in front of every step: it judges tool calls for intent and risk, scans tool outputs for instructions aimed at AI agents, and checks replies, mapping probabilities to allow, ask or warn.
Jev Agents & MCP#community#catalog-import#logicrw
Community
harshwasan/pi-jev-sentinelTypeScript
harshwasan/pi-jev-sentinelTypeScriptOpen-source Codex plugin that consults TypeSafe Jev about tools, models, tasks, skills, context, and strategies, diagnoses failed commands, and checks completion claims against evidence. The current package is version 0.4.0.
Jev Agents & MCP#community#catalog-import#logicrw
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.