jev-pilot
Python library adding a Jev-based System 1 layer to agents for intent routing, tool safety checks and stuck-loop detection.
Overview
jev-pilot is a PyPI library that gives autonomous agents a fast System 1 layer using TypeSafe's Jev, alongside a big LLM acting as System 2. The README targets agents built on Claude, GPT, Gemini, Llama, Hermes or custom runtimes, and positions Jev for split-second operational decisions such as intent routing, tool safety guardrails and breaking out of stuck agent loops, leaving deep reasoning and code generation to the LLM.
It cites about 0.3 seconds per answer and $42 per billion input tokens, attributing these to TypeSafe's published figures and live tests. The API key is resolved from an argument, environment variables or a user-only config file that can be saved once. A Persian translation of the documentation is included.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which intent or tool does this request map to?(inferred) | — |
| Yes / no | Is this tool action safe to run?(inferred) | yes · no |
| Yes / no | Is the agent stuck in a loop?(inferred) | yes · no |
How it works
- pip install jev-pilot and configure a key
- The agent calls jev-pilot before acting
- Jev returns probabilities for routing, safety or loop questions
- The agent proceeds, blocks or changes course
Strengths
- Simple drop-in for multiple agent runtimes
- Persisted key configuration with restricted permissions
- Keeps fast decisions out of the expensive LLM
Caveats
- Latency and cost figures come from TypeSafe's claims
- 'Zero hallucination' refers only to typed output
- Requires a TypeSafe key
Getting started
pip install jev-pilot, then JevPilot(api_key="...", save=True) once.At a glance
- Type
- Project
- Status
- beta
- Language
- Python
- Runtime
- Library
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Fast System-1 Decision, Arbitration & Safety Engine for Autonomous AI Agents (Powered by TypeSafe Jev)
logicrw/awesome-jev-projectsFast System-1 Decision, Arbitration & Safety Engine for Autonomous AI Agents (Powered by TypeSafe Jev).
hellogumbo/awesome-jev
Frequently asked questions
What does jev-pilot decide?
Fast operational questions like intent routing, tool safety and whether an agent is stuck.
Which agents does it work with?
The README names Claude, GPT, Gemini, Llama, Hermes and custom runtimes.
Where is my key stored?
Optionally in ~/.jev_pilot/config.json with user-only permissions.
Related projects
Plugin suite adding Jev judgments to DeepSeek Harness: tool pruning, dead-loop blocking, a safety guard and skill routing.
PluginbetaTypeScriptDeepSeek Harness plugin that registers a native jev_decide tool and declares its own policy for when the agent should ask Jev.
PluginbetaJavaScriptjevkit: zero-dependency CLI and eleven MCP tools giving agents typed Jev decisions for routing, triage, guarding, grep, rank and compaction.
MCP serverbetaJavaScriptPortable decision layer for agent harnesses: Jev picks from host-supplied capabilities while the host keeps permissions and execution.
ProjectbetaTypeScriptJev decision layer for agents: seven MCP tools, an embeddable library and escalate-only Claude Code hooks at harness boundaries.
MCP serverbetaTypeScriptLocal gateway, MCP server and Claude Code hooks that let coding agents delegate explicit choices, classification and scoring to Jev.
MCP serverbetaTypeScript