RunplaneAI Overview

Execution Governance Scenarios

AI Use Cases

Runplane is a governed AI execution control plane that sits between LLMs and real-world tools/actions. Below are common scenarios where Runplane governs AI agent execution. Each use case demonstrates how policy enforcement, risk scoring, and human approval workflows apply to real-world operations.

Note: Runplane governs execution, not reasoning. It is not an LLM, not an agent framework, and not a chatbot.

Payment Operations

Govern AI-initiated financial transactions

Policy Examples

  • •Block charges exceeding threshold without approval
  • •Require human approval for refunds over $500
  • •Allow balance inquiries without restriction
  • •Audit all payment mutations

Canonical Action Types

create_chargeprocess_refundupdate_subscriptionread_balance

Database Operations

Control AI access to data systems

Policy Examples

  • •Block DELETE operations on production tables
  • •Require approval for schema modifications
  • •Allow SELECT queries with rate limiting
  • •Log all write operations

Canonical Action Types

execute_querymodify_schemadelete_recordsbulk_update

Infrastructure Management

Govern AI-driven infrastructure changes

Policy Examples

  • •Block instance termination without approval
  • •Require approval for security group changes
  • •Allow read-only resource inspection
  • •Contain blast radius of deployments

Canonical Action Types

terminate_instancemodify_securitydeploy_servicescale_resources

Communication Systems

Control AI-initiated messaging

Policy Examples

  • •Block bulk email sends without approval
  • •Require approval for external communications
  • •Allow internal notifications
  • •Rate limit message volume

Canonical Action Types

send_emailsend_smspost_messagebroadcast_notification

Third-Party Integrations

Govern AI interactions with external APIs

Policy Examples

  • •Block API calls that modify external state
  • •Require approval for OAuth token actions
  • •Allow read-only data fetches
  • •Log all external API interactions

Canonical Action Types

call_apimodify_webhooksync_dataupdate_integration

Enterprise Workflows

Control AI in business processes

Policy Examples

  • •Require approval for contract modifications
  • •Block unauthorized data exports
  • •Allow document generation
  • •Audit compliance-sensitive operations

Canonical Action Types

generate_documentexport_datamodify_recordapprove_request

Integration Pattern

All use cases follow the same integration pattern using the guard() function:

const result = await runplane.guard(
  actionType,    // e.g., "create_charge"
  target,        // e.g., "stripe"
  context,       // e.g., { amount: 5000, currency: "usd" }
  executeFn      // Your actual tool execution function
);

Decision Outcomes

ALLOW

Action executes immediately. Tool function is invoked.

REQUIRE_APPROVAL

Execution pauses. Human approval required before proceeding.

BLOCK

Action prevented. Tool function is not invoked.

Related AI Documentation