jeprum 0.1.0

The live control room for AI agents — see everything, set rules, stop anything

Jeprum

The live control room for AI agents — see everything, set rules, stop anything.

What is Jeprum?

Jeprum is an open-source SDK that monitors and governs AI agents in real-time. It intercepts MCP tool calls, enforces guardrails (spending limits, tool blocking, rate limits), provides a kill switch, and ships telemetry to a cloud dashboard. Install in one command, integrate in 5 lines of code.

Quick Start

pip install jeprum

from jeprum import Jeprum

jp = Jeprum(api_key="jp_live_xxx", transport_mode="cloud")
monitored = jp.monitor(my_mcp_session, rules={
    "max_spend_per_day": 10.0,
    "blocked_tools": ["delete_*"]
})
result = await monitored.call_tool("search", {"query": "test"})

Features

• Real-time monitoring — every tool call captured with duration, cost, and I/O
• Guardrails — spending limits, tool blocking, rate limiting, alert patterns
• Kill switch — stop any agent remotely from the dashboard
• Cloud dashboard — live event timeline, cost tracking, agent management
• MCP-native — wraps ClientSession with zero agent code changes
• Non-blocking — telemetry shipping never slows your agent
• Open source — Apache 2.0

How It Works

Agent → Jeprum SDK (intercept + guardrails) → MCP Server → Tool
               ↓
        Jeprum Cloud (dashboard, kill switch, audit trail)

The SDK wraps your MCP session. Every call_tool() is intercepted: guardrails are checked synchronously (before execution), events are shipped asynchronously (after execution). Your agent code doesn't change.

Configuration

rules = {
    "max_spend_per_day": 10.0,
    "blocked_tools": ["delete_*", "drop_*"],
    "alert_on": ["payment.*", "transfer.*"],
    "rate_limit": {"max_events": 100, "period_seconds": 60}
}

Cloud Dashboard

Sign up at jeprum.com to get an API key. The dashboard provides:

• Live event timeline with guardrail status
• Agent list with cost and status tracking
• Kill / pause / resume controls
• Cost-over-time charts

Documentation

• Getting Started
• Guardrail Rules
• API Reference
• Contributing

Contributing

See CONTRIBUTING.md for development setup and guidelines.

License

Apache 2.0 — see LICENSE for details.

Built by Jithendra Siddartha, MS CS @ NYU Tandon.