agentgateway Jev guardrail example
Example of using Jev as a webhook guardrail inside the agentgateway proxy to inspect model requests and responses.
Overview
Agentgateway is an open-source proxy for agent-to-LLM, agent-to-tool and agent-to-agent traffic built on MCP and A2A. It offers an OpenAI-compatible LLM gateway with budgets and failover, an MCP gateway with tool federation, A2A support, inference routing, auth, RBAC with a CEL policy engine, rate limiting and OpenTelemetry tracing. Its guardrail layer supports regex, moderation services and custom webhooks.
This entry refers to an example that plugs Jev into that guardrail layer through a webhook, so requests and responses passing through the proxy are judged before they continue, with the proxy's tracing and cost tracking around it. The main README does not describe the Jev example itself; details come from the list summaries.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Should this model request or response be blocked (e.g. prompt injection or leaked secret)?(inferred) | yes · no |
How it works
- Traffic passes through agentgateway
- Guardrail webhook sends the prompt or response to Jev
- Jev returns a probability for the policy question
- Gateway allows or blocks and records the trace and cost
Strengths
- Reuses an existing production proxy with tracing and spend controls
- Guardrail runs centrally for every agent behind the gateway
Caveats
- The Jev example is not documented in the main README
- Requires running the full proxy
Getting started
See the standalone quickstart at agentgateway.dev/docs; the Jev guardrail is configured as a custom webhook.At a glance
- Type
- Proxy
- Status
- production
- Language
- Rust
- Runtime
- proxy (standalone or Kubernetes)
- Has code
- yes
- License
- Apache-2.0
Use cases and integrations
How curated lists describe it
A Jev guardrail example in Agentgateway using a webhook to inspect model requests and responses.
logicrw/awesome-jev-projectsJev as an LLM prompt guardrail inside the agentgateway proxy, with tracing and cost tracking.
hellogumbo/awesome-jev
Frequently asked questions
Where does Jev sit?
Behind agentgateway's custom webhook guardrail, judging requests and responses.
Is Jev built into agentgateway?
No, it is wired in through the generic webhook guardrail in an example.
What else does the proxy add?
Tracing, cost tracking, auth, RBAC and rate limiting.
Related projects
Hosted guardrails API checking LLM inputs, outputs and retrieved documents for injection, jailbreaks and leaks in one Jev call.
ServicebetaTypeScriptGitHub Action that asks Jev yes/no questions about each diff chunk for secrets, vulnerabilities, complexity and prompt injection.
ProjectbetaPythonJev Sentinel: guard for Pi, Claude Code and Codex that checks tool calls, tool outputs and replies, failing closed.
PluginbetaTypeScriptRespan's behaviour classifier for AI traces that returns present, absent or not observable for each plain-language behaviour you define.
ModelSplunk search command that asks Jev typed questions about events and adds probabilities, choices and scores as fields.
AppbetaPythonOpenlayer library packing agent evals and guardrails (tool choice, grounding, injection, PHI) into one decision-model request.
BenchmarkbetaPython