# Platform Capacities

### Agents

An **agent** is a dedicated execution environment within your organization. You can create multiple agents, and each agent can manage many playground and process automations.

| Parameter                       | Optimal Capacity      |
| ------------------------------- | --------------------- |
| **Agents** *(per organization)* | Up to 100 agents      |
| **Playgrounds** *(per agent)*   | Up to 200 playgrounds |
| **Processes** *(per agent)*     | Up to 200 processes   |

{% hint style="info" %}
Optimal capacity values are **recommendations**. You can create additional agents, playgrounds, and processes, but these ranges help ensure the best automation performance and user experience. See [What happens if I exceed the recommended numbers?](#id-3.-what-happens-if-i-exceed-the-recommended-number-of-agents-playgrounds-or-processes) for details.
{% endhint %}

### Automation Design

A **playground** or **process** automation should be written within the following capacity guidelines:

| Automation Length  | Performance Capacity                                            |
| ------------------ | --------------------------------------------------------------- |
| **1-200 lines**    | Great — Fastest execution and easiest to maintain               |
| **200-500 lines**  | High — Suitable for more complex workflows                      |
| **500-1200 lines** | Moderate — Can be optimized by breaking into subprocesses       |
| **1200+ lines**    | Needs Optimization — Should be broken into smaller subprocesses |

| Parameter           | Optimal Capacity |
| ------------------- | ---------------- |
| **Number of Loops** | Up to 200 loops  |

{% hint style="success" %}
Split large automations into subprocesses to improve performance and maintainability.
{% endhint %}

### Runs

A **run** is a single execution instance of a process. **Concurrent runs** refer to the total number of runs executing at the same time across your organization. Runs can be started manually, triggered via email or API, or scheduled. The platform can support the following number of concurrent runs:

| Parameter           | Supported Capacity      | Description                                                                                                                                                                                             |
| ------------------- | ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Concurrent Runs** | 150 runs                | The total number of actively running instances across your entire organization. This includes all process runs (you can run the same process multiple times), parallel child runs, and test suite runs. |
| **Parallel Runs**   | 50 runs *(per process)* | The maximum number of child processes that a parent process can call and run at the same time. These count toward the concurrent run capacity.                                                          |
| **Test Suite Runs** | 40 runs                 | The maximum number of test runs that can execute simultaneously in the Test Suite when validating process changes. Test runs also count toward the concurrent run capacity.                             |

{% hint style="info" %}
See [FAQ #4](#id-4.-whats-the-difference-between-concurrent-parallel-and-test-suite-runs) for more details.
{% endhint %}

### Learnings

Learnings are rules that guide your automation's decision-making. Learnings are isolated to each agent, with no limit on the total number of learnings per agent.

| Parameter           | Supported Capacity |
| ------------------- | ------------------ |
| **Learning Length** | 15,000 characters  |

{% hint style="success" %}
For optimal performance, we recommend keeping the total number of learnings under 500 per agent.
{% endhint %}

### Emails

Kognitos supports sending and receiving emails with attachments **up to 20 MB**. Outgoing emails with oversized attachments are handled using download links. These limits apply only to native email functionality and not to Books (e.g., Outlook, Gmail).

| Parameter      | Supported Capacity              |
| -------------- | ------------------------------- |
| **Email Size** | 20 MB *(including attachments)* |

{% hint style="info" %}
For larger input files, use the [REST API](https://docs.kognitos.com/legacy/legacy-experience/rest-api) to trigger new runs.
{% endhint %}

### Files

The platform supports uploading and processing the following file types:

| File Type     | Extensions                     | Supported Capacity                                                                                                                                                                                              |
| ------------- | ------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **PDF**       | .pdf                           | 1 GB                                                                                                                                                                                                            |
| **Images**    | .jpeg, .jpg, .png, .tif, .tiff | 1 GB                                                                                                                                                                                                            |
| **Data**      | .txt, .yml, .yaml, .csv        | 1 GB                                                                                                                                                                                                            |
| **Documents** | .docx, .doc                    | 1 GB                                                                                                                                                                                                            |
| **HTML**      | .html, .htm                    | 1 GB                                                                                                                                                                                                            |
| **Email**     | .eml, .msg                     | 1 GB                                                                                                                                                                                                            |
| **ERP**       | .edi                           | 1 GB                                                                                                                                                                                                            |
| **Media**     | .wav, .mp3, .mp4, .mov, .avi   | 1 GB                                                                                                                                                                                                            |
| **Other**     | .ttf                           | 1 GB                                                                                                                                                                                                            |
| **Excel**     | .xlsx, .xls, .xlsb, .xlsm      | <p>250 MB <em>(.xlsx and .xlsm)</em><br>150 MB <em>(.xlsb and .xls)</em></p>                                                                                                                                    |
| **JSON**      | .json                          | <p>1 GB\*<br><br>While uploads up to 1 GB are supported, files larger than <strong>100 KB</strong> can't be fully displayed in the interface. These can be downloaded to view the complete content locally.</p> |

### Extraction from Documents

When extracting data or tables from documents using [LLMs](https://docs.kognitos.com/llms), the maximum supported document size depends on the model provider:

| Provider   | Supported Capacity      |
| ---------- | ----------------------- |
| **Gemini** | 50 MB *(per document)*  |
| **OpenAI** | 512 MB *(per document)* |

{% hint style="info" %}
**Note**: Standard [file upload limits](#files) still apply.
{% endhint %}

{% hint style="success" %}
For large documents, consider processing them in smaller chunks (e.g., 50-100 pages at a time) to improve performance and reliability. You can use [subdocument extraction](https://docs.kognitos.com/legacy/legacy-experience/automation-areas/llm/automation-procedures/extract-subdocuments) to split documents into manageable sections.
{% endhint %}

### Department Boxes

Store structured information at scale with capacities designed for fast retrieval and updates.

| Parameter   | Supported Capacity             |
| ----------- | ------------------------------ |
| **Entries** | 150,000 *(per department box)* |

***

## Frequently Asked Questions

### 1. What if I need more capacity?

These capacities reflect current performance levels and will continue to expand as the platform grows. If your workload requires higher limits, please reach out to our support team via **in-app chat** *(preferred)* or email us at **<support@kognitos.com>**.

### 2. What's the difference between agents and automations?

**Agents** are dedicated execution environments that contain and manage your automations. Each agent is dedicated to your organization, with its own skills and settings, and can hold multiple automations.

**Automations** are sets of instructions that tell Kognitos what to do — process data, work with documents, send emails, call other processes, and more. These are workflows that can take the form of an early idea (**playground**) or a complete **process**. Both playgrounds and processes are considered automations.

{% @mermaid/diagram content="flowchart TB
subgraph AGENT\[**Agent**]
direction TB
subgraph AUTOS\[**Automations**]
direction TB
A1\["Process a Document"]
A2\["Billing Pipeline"]
A3\["Customer Onboarding"]
end
end

%% Styling for automations
style A1 fill:#F2FB90,stroke:#333,stroke-width:1px,color:#000
style A2 fill:#F2FB90,stroke:#333,stroke-width:1px,color:#000
style A3 fill:#F2FB90,stroke:#333,stroke-width:1px,color:#000

%% Outer containers
style AUTOS fill:#f5f5f5,stroke:#666,stroke-width:1px,color:#000
style AGENT fill:#fafafa,stroke:#444,stroke-width:1.5px,color:#000
" %}

### 3. What happens if I exceed the recommended number of agents, playgrounds, or processes?

The optimal capacity recommendations (100 agents, 200 playgrounds, 200 processes) are guidelines, not hard limits. You can create unlimited agents, playgrounds, and processes. However, exceeding these recommendations may slow down the user interface, causing delays when loading dashboards, navigating lists, or accessing settings. For the best experience, we recommend staying within these ranges.

### 4. What's the difference between concurrent, parallel, and test suite runs?

**Concurrent runs** refer to the total number of active runs across your entire organization. You can run the same process multiple times, and each run counts as one concurrent slot.

> Example: With a 150 concurrent run limit, you could run the same process 150 times, or 75 runs each of two different processes, or any other combination up to 150 total running instances.

**Parallel runs** refer to child processes that a single automation spawns to execute simultaneously. One parent process can spawn up to 50 parallel children at once. These parallel runs count toward your concurrent run capacity.

> Example: If one process spawns 50 parallel children, it uses 51 concurrent slots (1 parent + 50 children).

**Test suite runs** are special runs used to validate process changes by comparing results against expected outcomes. You can run up to 40 test runs simultaneously. Test suite runs also count toward the concurrent run capacity.

### 5. What types of runs count toward the concurrent run capacity?

Only **active runs** count toward the concurrent run capacity. Process runs do **NOT** count when:

* Waiting for user input (exceptions)
* Waiting to retry after an error
* Completed (successfully, failed, or stopped)

When a process pauses for user input, it releases capacity for other processes to run.

### 6. Does it matter how I start a run for it to count toward the capacity limit?

No. All runs count toward the run limit regardless of how they're started — whether triggered manually, via API, by email, or on a schedule.

### 7. What happens when I reach the concurrent run limit?

When you reach the capacity limit for concurrent runs:

* API calls to trigger new runs may return 403 errors or "Resource Not Available" messages
* Process crashes may occur and require reruns or retries
* Processes beyond the capacity threshold may experience failures

### 8. Can I retry a process that was rejected due to exceeding capacity?

Yes. If a process is rejected due to exceeding capacity, you can retry it once capacity becomes available. We recommend staying within the capacity thresholds to prevent unexpected behavior.

### 9. Are API requests limited?

Yes, API endpoints all share the same rate limit pool, including both trigger and query operations (such as status checks). Rate limits differ by region. To check your rate limits, go to **API Keys** → click on the menu (<kbd>⋮</kbd>) to the right of an API key → **Usage**.