PLAYGROUND
Try to get past obsvr
See how obsvr evaluates every AI agent action in real time, from interception to audit record.
Secure Your Agent
Compare native agent execution against obsvr-governed execution to see policy enforcement in action.
import OpenAI from 'openai'const client = new OpenAI()const res = await client.chat.completions.create({model: 'gpt-4o',messages: [{ role: 'user', content: userInput }],tools: [sqlTool, searchTool],})// tool call executed directly - no validationawait executeTool(res.tool_calls[0])
import OpenAI from 'openai'import { obsvr } from '@obsvr/sdk'// one-time setupobsvr.init({ environment: 'production' })const client = obsvr.wrap(new OpenAI(), { user_id: req.user.id })const res = await client.chat.completions.create({model: 'gpt-4o',messages: [{ role: 'user', content: userInput }],tools: [sqlTool, searchTool],})// tool call is intercepted, policy-checked, and audited firstawait executeTool(res.tool_calls[0])
Default Governance Policies
Every AI agent action is evaluated through configurable policy controls. No LLM inference required. Pure code-based enforcement.
PII Protection
Detects sensitive data in prompts, responses, and tool outputs. Block, redact, hash, or review before release.
Tool Governance
Restrict which tools agents can invoke. Prevent shell access, database writes, or external API calls outside policy.
Human Approval
Require manual approval before high-risk actions like payments, emails, production changes, or customer communications.
Model Routing
Force sensitive requests to approved models only. Block unsupported providers or fallback behavior.
Prompt Injection Defense
Detect jailbreaks, prompt injection, and malicious instructions before they reach your models or tools.
Compliance Controls
Enforce your organization's policy rules with configurable, versioned governance controls.
Cost Governance
Apply per-user, per-agent, or per-tenant token budgets and spending limits automatically.
Custom Rules
Define your own policy rules with YAML or JSON. Any check, any condition, any action.
No login required. This is a client-side simulation of obsvr's policy engine. The real engine adds about 45 microseconds for full governance.
12
Contact
Deploy a control plane for your LLM and agent stack in under five minutes.
Control what your LLMs and agents are allowed to do. See obsvr intercept, enforce, audit, and export evidence from real LLM and agent traffic.