> ## Documentation Index
> Fetch the complete documentation index at: https://manager.harmix.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Quotas and limits

> Memory MCP rate limits, paid plan entitlements, and what counts toward retrieve quotas.

Memory MCP enforces per-account limits on `retrieve_memory` calls. Access requires an active **paid plan** (Memory MCP subscription).

## What is counted

**Toward your monthly retrieve limit:** successful `tools/call` requests with `name: retrieve_memory` on `POST /v1/mcp/memory` — only when retrieval completes successfully.

**Toward the hourly burst limit:** every authenticated `retrieve_memory` attempt after entitlement checks (including blocked calls).

**Logged but not counted toward monthly quota:** blocked calls after authentication (quota exceeded, entitlement blocked, invalid arguments, etc.) appear in **[Request Log](https://pam.harmix.ai/request-log)** but do not decrement your monthly retrieve allowance.

**Not counted at all:**

* `initialize`, `initialized`, `tools/list`, `resources/list`, `prompts/list`
* Developer REST API calls (`/v1/dev/*`)

## Paid plan limits

Memory MCP is offered on a single **paid plan**. Typical limits:

| Limit                              | Value                        |
| ---------------------------------- | ---------------------------- |
| Successful `retrieve_memory` calls | **1,000** per billing period |
| Hourly burst                       | **100** calls per UTC hour   |
| Connected OAuth memory sources     | **1**                        |

Each successful retrieval uses one retrieve request from your monthly allowance. Handshake and discovery methods are free.

## Reset schedule

| Window                     | Resets at                                                                                                |
| -------------------------- | -------------------------------------------------------------------------------------------------------- |
| Hourly burst               | Start of each UTC hour                                                                                   |
| Monthly retrieve allowance | \~32 days after your last subscription payment (`reset_at` in the Usage UI and `GET /v1/dev/mcp-config`) |

## Entitlement gate

`entitlement_required` is returned when Memory MCP is not available on your account — for example, no active paid plan, Memory MCP not enabled, subscription inactive, or no chat credits remaining on your PAM account.

Upgrade or renew on **[Setup](https://pam.harmix.ai/setup)** to continue.

## Near-limit warnings

The **[Usage](https://pam.harmix.ai/usage)** page in the PAM app shows quota status:

| Status                 | Meaning                                  |
| ---------------------- | ---------------------------------------- |
| `ok`                   | Within limits                            |
| `near_limit`           | Above 80% of monthly retrieve limit      |
| `limited`              | Monthly retrieve limit reached           |
| `entitlement_required` | Memory MCP not available on your account |

Check live quota on **Usage**, or via `GET /v1/dev/mcp-config` (see [Setup and keys](/docs/setup-and-keys)).

## Example: quota\_exceeded response

Tool-level error shape (HTTP 200 with `isError: true`):

```json theme={null}
{
  "content": [
    {
      "type": "text",
      "text": "Rate limit exceeded. Please try again later."
    }
  ],
  "structuredContent": {
    "status": "error",
    "error_code": "quota_exceeded",
    "error_message": "Rate limit exceeded. Please try again later."
  },
  "isError": true
}
```

## Request history

Each `retrieve_memory` attempt stores the full prompt and full tool response for review on **[Request Log](https://pam.harmix.ai/request-log)** and via `GET /v1/dev/mcp-requests`.

<Warning>
  `quota_exceeded` applies only to `retrieve_memory` via `tools/call`, not to handshake methods or Developer REST calls.
</Warning>