---
title: Limits and pricing
description: What the Titan Harness costs in the private preview, how usage is
  billed, and the limits on tasks, requests, swarms, webhooks and evidence
  packs.
url: https://hexelstudio.com/docs/harness/limits-and-pricing
---

## Pricing

<Note>
Preview prices. These are the private preview's prices and may change before general availability.
</Note>

The Titan Harness is an add-on to a Titan plan. It is available on the current Pro, Business and Enterprise plans.

| Item | Price (preview) |
| --- | --- |
| Titan Harness add-on | $49 per organization per month, charged at the end of each billing period |
| Model input tokens (`harness.model_input_tokens`) | $3 per million |
| Model output tokens (`harness.model_output_tokens`) | $15 per million |
| Cached input tokens read (`harness.model_cached_tokens`) | $0.15 per million |
| Cache writes (`harness.model_cache_write_tokens`) | $3.75 per million |
| Hosted web searches (`harness.web_search_calls`), with web research on | $15 per 1,000 |
| Memory operations (`harness.memory_operations`), with agent memory on | $1 per 1,000 |
| Tasks created (`harness.tasks`) | No charge; counted |

- Hexel Studio attaches the add-on when it onboards your organization; it is not self-serve during the preview.
- Usage is charged to your organization's balance. The credit included with a Titan plan does not cover harness usage, so keep a balance: top up in the console. A task that the balance cannot cover is refused with `402 insufficient_balance`.
- During the preview, the dedicated Titan instance and runtime that Hexel Studio runs for your organization have no separate charge.
- Usage is attributed to the person who created each task, in the task's workspace and environment, and shows on your organization's costs.
- [Estimates](https://hexelstudio.com/docs/harness/budgets-and-estimates#estimates) price a task or a swarm with your organization's own prices before it starts.

## Limits

### Tasks

| Limit | Value |
| --- | --- |
| Tasks queued, running or waiting at once, per organization | 10 (`429 too_many_running_tasks`). A paused task does not count. |
| Open swarms per organization | 100 (`429 too_many_swarms`) |
| A swarm's members working at once | 1 to 100 (default 4) |
| Items in a `map` or `pipeline` swarm, or on a board | 1,000 |
| Tasks a chained task waits for | 20; `inputs_from` values: 32 |
| Check-in interval | 1 hour to 30 days |

### Requests

| Limit | Value |
| --- | --- |
| Request body | 1 MiB (`413 too_large`) |
| A task's `inputs` | 256 KiB |
| A task's `goal`, a steer's `text`, an answer | 64 KiB each |
| A decision's `reason` | 4 KiB |
| A spec bundle | 512 KiB |
| Content relayed live per event | 1 MiB (`truncated` on the last chunk) |
| Open event streams per organization | Limited (`429 too_many_streams`) |
| Artifact downloads waiting per organization | 20 (`429 too_many_downloads`) |
| A `shell` command | 280 seconds |

### Webhooks and evidence packs

| Limit | Value |
| --- | --- |
| Webhooks per environment | 20 |
| Webhook tests | One every 5 seconds per webhook; 10 a minute per organization |
| Webhook delivery attempt | 10 seconds; deliveries kept 7 days |
| Evidence pack period | 31 days and 200 tasks |
| Evidence pack size | 200,000 Ledger entries or 256 MiB (`422 evidence_pack_too_large`) |
| Evidence packs | One build at a time per organization (`429 evidence_pack_busy`); 200 requests per organization a day |

### Data

| Limit | Value |
| --- | --- |
| Retention | 30 days by default; 1 to 3,650 days ([Data retention](https://hexelstudio.com/docs/harness/data-retention)) |
| Replay page | 1 to 1,000 events |
