Jev is TypeSafe AI's first public System One model. The Jev AI model evaluates structured questions and returns typed choices, probabilities and scores instead of generating long-form text. AutoJev packages Jev for real agent workflows.

Is the Jev model a chatbot or generative LLM?

No. The Jev model is designed for bounded decisions inside software: classify, route, score, verify or choose among options defined by the caller. Use a generative model for writing and planning, then use Jev at decision boundaries.

Jev Skills are portable agent instructions that describe when to call a Jev MCP tool, which task state to send and how to apply the typed result. AutoJev publishes one umbrella tool router plus focused Skills for model routing, task routing, tool-call protection, research checks and completion review.

Why not call the Jev endpoint directly?

You can, but an endpoint alone does not give an agent stable schemas, tool discovery, trigger instructions, access control, or a policy for acting on the result. AutoJev supplies that operational layer through six MCP tools, one umbrella router Skill, five workflow Skills, REST presets, and deterministic guidance.

Why offer both MCP and skills?

MCP exposes callable tools and their schemas. Skills teach an agent when a decision boundary has been reached, what context to send, and how to handle the typed result. REST supports workflows that do not use MCP.

Can AutoJev execute actions on its own?

No. Decision packs return structured answers, probabilities, and preset guidance. Your agent or workflow decides what happens next, within its existing permissions.

What context will be sent to Jev?

Each tool uses a documented input schema so the agent can send compact, task-specific state. Do not include secrets, provider keys, private customer data, or unrelated files.

Is AutoJev available now?

Yes. The first release includes a remote MCP server with six tools, one umbrella router Skill, five portable workflow Skills, preset REST endpoints, and a generic Decisions API.

Jev: typed AI decisions for every agent.

AutoJev connects Jev, TypeSafe AI's decision model, to AI agents through a hosted MCP server, typed API and installable Skills for routing, guardrails and verification.

Open live playground Read integration docs

Connect AutoJev

https://autojev.ai

Add the Jev MCP server

{
  "mcpServers": {
    "autojev": {
      "type": "http",
      "url": "https://autojev.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${AUTOJEV_API_KEY}"
      }
    }
  }
}

Agent setup

Point your MCP client at the hosted AutoJev endpoint shown here.

Create an access key

Jev in practice

Put Jev to work routing Codex, Claude Code and your other agents.

AutoJev is a free, open-source desktop gateway. Its intelligent selection asks Jev which model fits each request — using routing metadata only, never your conversation.

Download AutoJev How Jev routing works

Jev model guide

The Jev AI model decides. Generative models create.

Jev is TypeSafe AI's System One model for typed decisions inside software. Give the Jev model shared state and bounded questions, then receive choices, probabilities and scores instead of prose.

Read the complete Jev guide

TypeSafe Jev guide Jev AI for agents Jev model guide Jev API guide Jev ecosystem Jev Skills

DimensionJevGenerative LLM

OutputTyped choices, probabilities, and scoresGenerated text, code, and tool arguments

Best useRouting, classification, scoring, and gatesPlanning, explanation, coding, and creation

UncertaintyCalibrated probabilities are part of the responseConfidence handling depends on the workflow

Jev MCP

Jev MCP turns a decision model into tools an agent can discover.

AutoJev adds typed schemas, authentication, stable tool names, preset guidance, and portable Skills around Jev. The agent keeps its existing model, tools, permissions, and approval rules.

Explore Jev MCP

Decisions your agent should not make from instinct alone.

AutoJev packages Jev into typed control points that agents can discover, call, and apply consistently.

autojev_route_model

Smart model routing

Choose from the models you allow using task stakes, quality, cost, latency, context, and tool-use needs. Escalate when no candidate is safe enough.

Input: task + candidates
Signals: quality / cost / latency
Output: model + escalate

autojev_guard_tool_call

Tool-call protection

Inspect the proposed action, side effects, safeguards, reversibility, and your policy before an agent changes external state.

allow / confirm / review / deny

Research Guard

Judge whether the available evidence is sufficient, contradictory, stale, or ready to cite.

Completion Reviewer

Decide whether a task is truly done, needs one more check, or should return to implementation.

Custom decision contracts

Define your own choice, yes-or-unknown, and scored questions through the generic Decisions API.

One Jev decision layer, four ways to use it.

AutoJev provides the connection, contracts, and workflow instructions needed to move Jev decisions from a model endpoint into real agent runs.

Jev MCP server

https://autojev.ai/mcp

Expose six typed Jev tools to Codex, Claude Code, Cursor, and any Streamable HTTP MCP client.

Configure MCP

Decisions REST API

https://autojev.ai/api/v1/decisions/{preset}

Call model routing, tool protection, research, task routing, completion, or custom decision contracts from any backend.

View API docs

Jev Skills for AI agents

https://autojev.ai/skills/autojev-*/SKILL.md

Install portable Jev Skills that teach an agent when to call each decision tool, what state to send, and how to apply the typed result without expanding authority.

Install Skills

Online Jev playground

https://autojev.ai/playground

Edit real preset inputs, call the configured AutoJev API, and inspect choices, probabilities, scores, usage, and guidance.

Start debugging

Live Jev API debugger

Inspect the decision before your agent acts.

Start with an agent-ready AutoJev workflow preset, or use the Jev Decisions contract for a custom decision. Edit the real JSON request and inspect exactly what the service returns. Your AutoJev key stays in browser memory and is never saved.

Recommended for agents: call this stable AutoJev workflow schema directly through MCP or REST. AutoJev validates the business fields, converts them into Jev state + questions, and returns deterministic action guidance with the raw answers.

AutoJev API key

Used only for this request. The page does not persist the key.

Request endpoint

/api/v1/decisions/tool-guard

The debugger sends this relative path to the current AutoJev service.

AutoJev preset input

{
  "tool": "issue_customer_refund",
  "action": "Refund USD 680 after a disputed duplicate charge",
  "arguments_summary": [
    "order_id=ord_7429",
    "amount_usd=680"
  ],
  "side_effects": [
    "Moves funds",
    "Changes the order payment state"
  ],
  "safeguards": [
    "Customer identity and duplicate charge verified"
  ],
  "policy": [
    "Refunds above USD 500 require human approval"
  ],
  "reversibility": "partially_reversible"
}

Create an access key

Structured response

decision + guidance + answers

The typed Jev response will appear here.

Run the request to inspect Jev answers, confidence, probability distributions, model version, token usage, and any deterministic AutoJev preset guidance.

Decision boundaries

Use Jev where the workflow needs an answer, not another paragraph.

Each AutoJev use case turns compact workflow state into a bounded decision your agent can inspect and apply within its existing permissions.

Explore all use cases

Model routing

Choose the model that fits the task

Compare only the models you allow against quality, context, tool use, cost, latency and task stakes.

Workflow question: Which approved model can handle 100k context and tool use without wasting cost?
Example decision: reasoning-model

Read the use case Try this preset

Tool-call guardrails

Review an action before it changes external state

Evaluate side effects, safeguards, policy thresholds and reversibility before an agent calls a consequential tool.

Workflow question: Can the agent issue this high-value refund without human confirmation?
Example decision: review · 0.91

Read the use case Try this preset

Task routing

Select the next workflow path

Route ambiguous work to quick execution, deeper review, decomposition or a blocked state using explicit evidence.

Workflow question: Should the agent proceed, review deeply, split the task or stop?
Example decision: review_deep

Read the use case Try this preset

Research verification

Check whether the evidence supports the claim

Judge evidence coverage, contradictions, source quality and stakes before an agent cites or acts on research.

Workflow question: Is the current evidence sufficient to support this policy claim?
Example decision: insufficient

Read the use case Try this preset

Completion review

Keep unfinished work from being reported as done

Compare the objective, completed work, verification and known gaps before closing the task.

Workflow question: Is the refund task complete if the charge is identified but no refund was issued?
Example decision: verify_more

Read the use case Try this preset

The value is not another model endpoint.

Jev supplies fast, calibrated judgment. AutoJev turns that judgment into a control surface an agent can safely discover and use.

Decision engine

Jev evaluates

A System One model selects among explicit options and returns probabilities and scores.

AutoJev

Operational decision layer

Contract

Typed inputs and bounded outputs for each decision.

Connect

MCP tools, REST presets, and portable skills.

Trigger

Teach agents which boundary needs a second judgment.

Action policy

Map an answer to proceed, confirm, review, or stop.

Execution layer

Your agent acts

The agent keeps its model, tools, permissions, and human approval boundaries.

Agent integration

Connect once. Add judgment at every decision boundary.

MCP makes the tools discoverable. Skills teach when to call them. Typed results give the workflow a predictable next action.

Connect the MCP server

Add one remote endpoint and keep the AutoJev access key in an environment variable.

https://autojev.ai/mcp

Install the right skills

Give the agent explicit triggers for routing, tool protection, research, and completion checks.

.agents/skills/autojev-*

Evaluate with Jev

Send compact task state and bounded questions, then receive choices, probabilities, and scores.

choice + noul + score

Apply your policy

Let the workflow proceed, request confirmation, route for review, or stop without expanding authority.

allow / confirm / review / deny

Make the first guarded decision.

The agent discovers six typed tools over MCP. Add portable skills when you want it to call those tools at specific workflow boundaries.

Create an AutoJev access key from your account settings.
Point your MCP client at the hosted AutoJev endpoint shown here.
Install all six AutoJev Skills in one step.

Create an access key Read the integration guide

Remote MCP configuration

Codex / config.toml

[mcp_servers.autojev]
url = "https://autojev.ai/mcp"
bearer_token_env_var = "AUTOJEV_API_KEY"

Streamable HTTP MCP

Bearer API key

Typed Jev output

Jev AI for Agents: Typed Decisions, Routing and Guardrails

Learn how Jev AI gives software and AI agents typed decisions for classification, routing, scoring, guardrails and workflow verification.

Jev AI is a decision model from TypeSafe AI built for structured judgments inside software. Give the Jev model state plus bounded questions, and it returns typed choices, probabilities and scores that application code can use directly.

This page covers the AI model named Jev. It is unrelated to the medical abbreviation JEV commonly used for Japanese encephalitis virus.

Jev AI in one minute

Question	Short answer
What is Jev AI?	TypeSafe AI's System One model for bounded decisions inside software
What does it return?	Typed choices, scores, probabilities and confidence—not open-ended prose
Where does it fit?	Between generative model output and application-owned policy at routing, guardrail and review steps
What remains in code?	Permissions, hard rules, thresholds, side effects and human-approval requirements
How can I try it?	Use the Jev playground, Jev API or Jev MCP server

What does Jev AI do?

Jev AI evaluates information without writing an open-ended response. The caller defines the question and the permitted answer space before the request. Jev then estimates which answer best fits the supplied state and communicates uncertainty through probabilities and confidence.

That makes Jev useful for decisions such as:

Classify an inbound support request.
Route a task to one of several available AI models.
Score the risk of a proposed tool call or deployment.
Check whether research evidence supports a claim.
Decide whether an agent task is complete.
Estimate whether a human should review the next action.

Jev does not send the support reply, run the selected model, execute the tool or deploy the release. Your application or agent remains responsible for actions, permissions and approval boundaries.

Where Jev fits in an AI agent

A production agent usually needs more than one kind of intelligence:

A generative model reads instructions, plans, writes code or content and proposes tool calls.
Jev AI evaluates bounded questions at important workflow boundaries.
Deterministic code validates inputs, enforces permissions, applies thresholds and chooses the permitted next action.
A human reviewer handles low-confidence, irreversible or high-impact cases when policy requires it.

This division keeps each component focused. The generative model remains flexible, Jev supplies a typed judgment, and ordinary code retains control over side effects.

Jev AI compared with a generative LLM

Question	Jev AI	Generative LLM
Primary output	Typed decisions and probabilities	Text, code, plans and tool arguments
Available answers	Defined by the caller	Open-ended
Strong use cases	Classify, route, score, verify and gate	Create, explain, plan and transform
Output parsing	Known schema	Usually requires validation
Authority to act	None	None unless the host grants it
Can the judgment still be wrong?	Yes	Yes

Jev is not a replacement for every LLM call. It is a specialized model for the many points in software where the desired output is a decision rather than a paragraph.

Common Jev AI patterns

Model routing

Give Jev a task, the models your environment can actually call and the constraints that matter. Jev can choose among those candidates while also estimating whether the choice should be escalated for review.

Tool-call guardrails

Evaluate the proposed action, summarized arguments, side effects, safeguards, reversibility and policy immediately before a consequential tool call. Application code can map the result to allow, confirm, review or deny.

Research verification

Provide a precise claim plus compact evidence entries. Jev can classify the evidence as sufficient, needing more verification or unsuitable for the claim. Source retrieval and citation remain the host application's job.

Completion review

Compare the original objective with completed work, tests, failures and known gaps. Jev can help an agent decide whether to finish, verify more or return to implementation.

How AutoJev exposes Jev AI

AutoJev packages these patterns into interfaces an agent or application can use consistently:

A hosted Jev MCP server exposes discoverable, typed tools.
A Decisions REST API serves presets and custom Choice, Noul and Score questions.
One umbrella router and five focused Jev Skills teach agents when to call each tool.
An online Jev playground shows real request and response shapes.
The integration guide covers authentication and client configuration.

The hosted AutoJev service keeps its upstream model credential on the server. Client agents authenticate with an AutoJev access key and receive decisions, probabilities, usage metadata and deterministic preset guidance.

Jev AI limitations and safe use

A typed output is easier for software to consume, but it is not proof that the judgment is correct. Treat Jev probabilities as decision evidence, not as permission or certainty.

Evaluate important use cases against your own representative data.
Define an explicit low-confidence or escalation path.
Keep calculations, permission checks and hard business rules in code.
Preserve human approval for irreversible or high-impact actions.
Send only the state needed for the current decision.
Never put secrets, passwords or provider credentials into decision context.

Jev AI FAQ

Who makes Jev AI?

Jev is TypeSafe AI's first public System One model. TypeSafe describes System One models as models optimized for fast, structured decisions that software can use directly.

Does Jev AI generate text?

No. Jev is designed to return bounded, typed answers and probabilities. Use a generative model when the output must be prose, code, a plan or an explanation.

Can Jev AI control an agent?

Not by itself. Jev returns a decision. The surrounding agent, MCP host or application decides whether and how to use it within existing permissions.

How can I try Jev with an agent?

Use the AutoJev Playground to inspect presets, then connect the Jev MCP server and install all AutoJev Skills in one step.

Continue learning about Jev

Jev AI questions, answered.

Understand the Jev model, Jev MCP and Jev Skills before adding typed decisions to an agent workflow.

Give your agent a decision boundary.

Create an access key, connect the MCP server, and place typed judgment before high-impact actions.

Create an access key

Jev: typed AI decisions for every agent.

The Jev AI model decides. Generative models create.

Jev MCP turns a decision model into tools an agent can discover.

Decisions your agent should not make from instinct alone.

Smart model routing

Tool-call protection

Research Guard

Completion Reviewer

Custom decision contracts

One Jev decision layer, four ways to use it.

Jev MCP server

Decisions REST API

Jev Skills for AI agents

Online Jev playground

Inspect the decision before your agent acts.

Use Jev where the workflow needs an answer, not another paragraph.

Choose the model that fits the task

Review an action before it changes external state

Select the next workflow path

Check whether the evidence supports the claim

Keep unfinished work from being reported as done

The value is not another model endpoint.

Jev evaluates

Operational decision layer

Your agent acts

Connect once. Add judgment at every decision boundary.

Connect the MCP server

Install the right skills

Evaluate with Jev

Apply your policy

Make the first guarded decision.

Jev AI for Agents: Typed Decisions, Routing and Guardrails

Jev AI in one minute

What does Jev AI do?

Where Jev fits in an AI agent

Jev AI compared with a generative LLM

Common Jev AI patterns

Model routing

Tool-call guardrails

Research verification

Completion review

How AutoJev exposes Jev AI

Jev AI limitations and safe use

Jev AI FAQ

Who makes Jev AI?

Does Jev AI generate text?

Can Jev AI control an agent?

How can I try Jev with an agent?

Continue learning about Jev

Jev AI questions, answered.

What is Jev?

Is the Jev model a chatbot or generative LLM?

What are Jev Skills?

Why not call the Jev endpoint directly?

Why offer both MCP and skills?

Can AutoJev execute actions on its own?

What context will be sent to Jev?

Is AutoJev available now?

Give your agent a decision boundary.