Guard MCP API

Rate limiting and cost control for AI agents via Model Context Protocol

Overview

The Guard MCP API provides AI assistants with rate limiting and cost control capabilities. Optimized for AI integrations using the MCP protocol over HTTP transport with JSON-RPC 2.0.

MCP Endpoint

https://mcp.ainoflow.io/mcp/v1/guard

Transport:

HTTP (SSE)

Configuration

OpenAI Agent Builder

Add MCP server in the UI dialog:

URL:

Label:

Authentication (Access token / API Key):

Claude Desktop

Add to your Claude Desktop configuration:

Cursor IDE

Add to your Cursor MCP configuration:

Available Tools

guard_check

Check rate limit and increment execution counter. On first call with rateMax/rateWindow, a policy is auto-created.

key

required

- Guard key identifier (max 512 chars)

rateMax

optional*

- Max executions per window (1–1,000,000)

rateWindow

optional*

- Window size in seconds (1–86,400)

failOpen

optional

- Allow execution when the service is temporarily unavailable (default: false)

allowPolicyOverwrite

optional

- Update stored policy when rateMax/rateWindow differ instead of returning error (default: false)

* Required on first call to auto-create policy. Optional on subsequent calls.

Example Usage (first call):

Allowed Response:

Denied Response:

Check allowed field to determine if execution should proceed

guard_status

Check current counter status without incrementing. Uses stored policies.

key

required

- Guard key identifier

Example:

Response:

guard_reset

Reset rate limit counters for a key. For debugging only.

key

required

- Guard key identifier

Example:

Response:

Available Resources

guard://{key}

Read-only resource for checking counter status

MIME Type: application/json

Returns: {"key":"send-emails","exists":true,"rateLimit":{"current":42,"max":100,"window":3600,"resetsIn":2145}}

Use Cases for AI Assistants

Rate-Limited API Calls

Prevent expensive API overuse

Cost Budget Enforcement

Cap expensive operations per time window

Combined with Shield

Rate limit + deduplication

Check Before Batch

Check remaining budget before starting

Constraints and Limits

Key specifications for Guard operations

Key lengthMax 512 characters

rateMax range1 to 1,000,000

rateWindow range1 to 86,400 seconds

AlgorithmFixed window

Policy cache TTL5 minutes

Denied requestsNot counted toward limit

REST API Compatibility

Both MCP and REST APIs can be used simultaneously

REST API	MCP Tool
`POST /guard/{key}/counter`	`guard_check`
`GET /guard/{key}/counter`	`guard_status`
`DELETE /guard/{key}/counter`	`guard_reset`

View full REST API Documentation →

Ready to integrate Guard MCP?

Get Started Free Back to MCP Hub