惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Blog — PlanetScale
Blog — PlanetScale
Y
Y Combinator Blog
G
Google Developers Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
L
LangChain Blog
S
SegmentFault 最新的问题
J
Java Code Geeks
V
Visual Studio Blog
H
Help Net Security
Stack Overflow Blog
Stack Overflow Blog
aimingoo的专栏
aimingoo的专栏
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
T
Tailwind CSS Blog
Microsoft Azure Blog
Microsoft Azure Blog
博客园_首页
H
Hackread – Cybersecurity News, Data Breaches, AI and More
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
博客园 - Franky
B
Blog RSS Feed
The Cloudflare Blog
MyScale Blog
MyScale Blog
月光博客
月光博客
Microsoft Security Blog
Microsoft Security Blog
美团技术团队

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
AI Has Hands Now. You Should Decide What It Can Do.
HaveAGoodOne · 2026-05-04 · via DEV Community

We’ve crossed a line.

AI is no longer just generating text.

It’s starting to take actions.

Your agent can now:

  • call APIs
  • delete data
  • send emails
  • trigger workflows
  • modify production systems

And in most setups today, it does all of that without a control layer.

The uncomfortable truth

Most AI agents today operate like this:

await deleteUser(userId);
await sendEmail(customer);
await transferFunds(amount);

Enter fullscreen mode Exit fullscreen mode

If the agent decides to do it — it just… runs.

No checkpoint.
No approval.
No policy enforcement.

That’s fine in a demo.

It’s not fine in production.

This is where things break

The moment your AI touches real systems, you’re exposed to:

  • Accidental destructive actions
  • Prompt injection leading to unintended behavior
  • Over-permissioned tools
  • No audit trail of decisions

You don’t need a malicious AI.

You just need:

a slightly wrong decision, at the wrong time, in the wrong environment

The missing layer: decision before execution

What’s missing is simple:

Every action should be checked before it runs.

Not after.
Not in logs.
Not in alerts.

Before execution.

Introducing Runplane

Runplane adds a runtime control layer between your AI and real-world actions.

Instead of executing immediately, every action goes through a decision:

  • ✅ ALLOW
  • ❌ BLOCK
  • ⏸ REQUIRE APPROVAL

What this looks like in practice

Without control:

await deleteUser(userId);

Enter fullscreen mode Exit fullscreen mode

With Runplane:

await runplane.guard(
  "delete_user",
  "production-db",
  { userId },
  async () => {
    return deleteUser(userId);
  }
);

Enter fullscreen mode Exit fullscreen mode

Now that action can:

  • be blocked
  • require human approval
  • be logged and audited

This is not just theory

If you’re building:

  • AI agents
  • automation workflows
  • MCP-based tools
  • internal copilots
  • API-triggering systems

You already have the problem.

You just haven’t felt it yet.

Built for modern AI architectures (including MCP)

Runplane fits directly into current AI stacks:

  • Works alongside agent frameworks
  • Supports MCP-style tool execution flows
  • Sits between the model and the tool call
  • Does not require changing how your tools work

It doesn’t replace your agent.

It controls what your agent is allowed to do.

Why this matters now

We’re moving from:

“AI suggests actions”

to:

“AI executes actions”

That shift changes everything.

Execution without control is risk.

Try it (free developer tier)

We opened a free developer tier so you can test this in real flows:

https://runplane.ai/auth/sign-up?mode=developer

You can:

  • integrate in minutes
  • simulate risky actions
  • test approval flows
  • see exactly how decisions affect execution

Final thought

You don’t need to wait for a disaster to add control.

By the time something breaks, it’s already too late.

AI has hands now.

You should decide what it’s allowed to touch.