MeshGuard for Small Business: Governing AI Agents Without a Security Team

You don't need a CISO or a security operations center to govern your AI agents. MeshGuard gives small teams enterprise-grade controls with zero overhead.

The Problem: AI Agents Are Everywhere, Security Teams Are Not

Small businesses are adopting AI agents fast. A customer support bot here, an inventory assistant there, maybe an AI that handles email or schedules appointments. These agents are genuinely useful — but they come with real risk.

And here's the uncomfortable truth: most small businesses have no one responsible for AI governance. No security team. No compliance officer. No one watching what the agent does at 2 AM on a Saturday.

Why Governance Matters for a 3-Person Team

"We're only three people — what's the worst that could happen?"

Plenty:

• A customer service agent promises a full refund your policy doesn't support — and you have to honor it
• An AI email assistant sends the wrong attachment with another client's confidential data
• An ordering agent goes haywire and places 500 units instead of 5, costing you thousands
• A research agent scrapes and stores data that violates privacy regulations

One rogue agent action can cost more than your monthly revenue. Governance isn't bureaucracy — it's insurance.

MeshGuard's Free Tier: Built for Small Teams

MeshGuard's Free tier includes 5 governed agents — more than enough for most small businesses. No credit card required. No time limit.

Feature	Free Tier
Governed agents	5
Policy checks	1,000/month
Audit log retention	7 days
Support	Community
Price	$0

Real-World Scenarios

🛒 E-Commerce: Governing Customer Support

A small online shop uses an AI agent to handle customer inquiries — order status, returns, basic troubleshooting.

Risks without governance:

• Agent promises refunds beyond your return policy
• Agent shares other customers' order details
• Agent agrees to custom deals or discounts it shouldn't

MeshGuard policy:

name: ecommerce-support-agent
version: 1
agent_match:
  tags: ["customer-support"]

rules:
  # Can look up orders
  - action: "read:orders"
    effect: allow

  # Can process refunds up to $25
  - action: "write:refund"
    effect: allow
    conditions:
      - "request.amount <= 25"

  # Block refunds over $25 — escalate to owner
  - action: "write:refund"
    effect: deny
    conditions:
      - "request.amount > 25"
    reason: "Refunds over $25 require owner approval"

  # Cannot offer custom discounts
  - action: "write:discount"
    effect: deny
    reason: "Custom discounts are not available through automated support"

  # Rate limit customer emails to prevent spam
  - action: "write:email"
    effect: allow
    conditions:
      - "rate_limit(30, '1h')"

💼 Consulting Firm: Protecting Client Data

A 5-person consulting firm uses an AI research agent to gather information, draft reports, and summarize documents.

Risks without governance:

• Agent accesses Client A's data while working on Client B's project
• Agent sends research externally (email, API calls) containing proprietary info
• Agent stores sensitive data in unencrypted locations

MeshGuard policy:

name: research-agent-policy
version: 1
agent_match:
  tags: ["research"]

rules:
  # Can read public data sources
  - action: "read:web"
    effect: allow

  # Can read project files for the assigned client only
  - action: "read:client_files"
    effect: allow
    conditions:
      - "request.client_id == agent.assigned_client"

  # Cannot access other clients' files
  - action: "read:client_files"
    effect: deny
    reason: "Access restricted to assigned client data only"

  # Cannot send data externally
  - action: "write:external_api"
    effect: deny
    reason: "External data transmission is not permitted"

  # Cannot write to shared storage
  - action: "write:shared_drive"
    effect: deny
    reason: "Write access to shared storage requires manual approval"

🍕 Restaurant: AI for Reservations and Ordering

A restaurant uses an AI agent to handle phone reservations, online ordering, and basic customer questions.

Risks without governance:

• Agent accepts reservations for dates you're closed
• Agent takes orders for items that are 86'd (out of stock)
• Agent processes an unreasonable number of orders (bot attack)

MeshGuard policy:

name: restaurant-agent-policy
version: 1
agent_match:
  tags: ["restaurant-front"]

rules:
  # Can read menu and availability
  - action: "read:menu"
    effect: allow
  - action: "read:availability"
    effect: allow

  # Can create reservations during operating hours
  - action: "write:reservation"
    effect: allow
    conditions:
      - "request.party_size <= 10"
      - "request.date IN available_dates"

  # Block oversized parties — needs manager
  - action: "write:reservation"
    effect: deny
    conditions:
      - "request.party_size > 10"
    reason: "Large party reservations require manager confirmation"

  # Rate limit orders to prevent abuse
  - action: "write:order"
    effect: allow
    conditions:
      - "rate_limit(60, '1h')"
      - "request.total <= 500"

  # Block abnormally large orders
  - action: "write:order"
    effect: deny
    conditions:
      - "request.total > 500"
    reason: "Large orders require phone confirmation"

Getting Started in 10 Minutes

Step 1: Sign Up

Go to meshguard.app and create a free account. No credit card needed.

Step 2: Install the SDK

pip install meshguard

Step 3: Get Your Agent Token

In the MeshGuard dashboard, create a new agent and copy its token.

Step 4: Add Governance to Your Agent

from meshguard import MeshGuardClient

client = MeshGuardClient(
    gateway_url="https://dashboard.meshguard.app",
    agent_token="your-agent-token",
)

# Before any sensitive action, check with MeshGuard
def handle_refund(order_id: str, amount: float):
    decision = client.check(
        action="write:refund",
        context={"order_id": order_id, "amount": amount}
    )
    if not decision.allowed:
        return f"Cannot process refund: {decision.reason}"
    
    return process_refund(order_id, amount)

Step 5: Create Your First Policy

In the dashboard, create a policy (or use one of the templates above) and attach it to your agent. That's it — your agent is now governed.

Cost: Free for Most Small Businesses

The Free tier (5 agents, 1,000 checks/month) covers the vast majority of small business use cases. A typical customer support agent might make 20-30 policy checks per day — well within the free limit.

If you grow beyond that, the Starter tier at $29/month gives you 25 agents and 10,000 checks — enough for a growing team with multiple agents.

Tier	Agents	Checks/Month	Price
Free	5	1,000	$0
Starter	25	10,000	$29/mo
Pro	100	100,000	$99/mo

When to Upgrade

You've outgrown the Free tier when:

• You have more than 5 agents
• You need more than 7 days of audit log retention
• You want priority support or SLA guarantees
• You need team collaboration features (multiple admins)

The jump to Starter is painless — same policies, same SDK, just more capacity.

Next Steps

• MeshGuard Pricing — Compare all tiers
• MeshGuard Documentation — Full reference
• What is Agent Governance? — Understand the fundamentals
• Building a Customer Service Agent — Detailed tutorial with full code
• Governing Clawdbot Agents — If you're using Clawdbot as your agent platform
• Personal vs Enterprise Governance — How governance needs differ by scale

---

Start Free

Sign up at meshguard.app — 5 agents, 1,000 checks/month, no credit card. Your AI agents will thank you.