Plans include a monthly allowance, metered by steps and always-on connections. Run agents on PAREL-hosted models at the provider's price, with no markup, or on your own model keys.
# Plans
| Plan | Price | Included each month | Always-on connections | Members | Beyond the allowance |
|---|
| Free | $0 | 5,000 steps + $1 of hosted models | none (webhook channels only) | 1 | stops until next month |
| Hobby | $5 / month | $5 of usage | up to 3 | 3 | next invoice, up to $20 by default |
| Pro | $20 / month | $20 of usage | unlimited | unlimited | next invoice, up to $100 by default |
Free
- Price
- $0
- Included each month
- 5,000 steps + $1 of hosted models
- Always-on connections
- none (webhook channels only)
- Members
- 1
- Beyond the allowance
- stops until next month
Hobby
- Price
- $5 / month
- Included each month
- $5 of usage
- Always-on connections
- up to 3
- Members
- 3
- Beyond the allowance
- next invoice, up to $20 by default
Pro
- Price
- $20 / month
- Included each month
- $20 of usage
- Always-on connections
- unlimited
- Members
- unlimited
- Beyond the allowance
- next invoice, up to $100 by default
Need SSO, audit logs or volume pricing? Contact us.
# How usage is metered
- Step — $0.50 per 1,000. One model call the runtime made for your agent and received usage for, including sub-agent and plugin calls. Failed calls and retries are not counted.
- Always-on connection — $5 per month, counted per day. An enabled long-lived connection (e.g. Slack Socket Mode). Webhook channels are free.
- Allowances reset on the 1st of each month (UTC); unused allowance does not roll over.
- On Hobby and Pro, usage beyond the allowance is added to the next monthly invoice. A spend cap (adjustable in the console) bounds it: at the cap, new turns pause until next month or until you raise it.
- Upgrades apply immediately and are prorated; downgrades and cancellations take effect at the end of the billing period.
# Models: hosted, or your own key
- PAREL-hosted models: a Claude or GPT model you give no key runs on PAREL's, and
provider: parel picks any other model PAREL hosts (Gemini, DeepSeek and more). Each call is billed at the price the model provider charges, with no markup, and counts toward your plan's allowance. The Free plan includes $1 of hosted models a month. - Your own key: add a provider key — Anthropic, OpenAI, OpenRouter, or any OpenAI- or Anthropic-compatible endpoint — and that agent's calls run on it. That provider bills you directly, and PAREL charges only the steps.
- One-click OpenRouter: connect your OpenRouter account in the console to use one key across many models.
- The console shows model spend per agent and per day: the actual cost for hosted calls, an estimate for calls on your own keys.
# Plugin secrets
Runtime plugins keep their own secrets separately. For example, E2B is referenced in agent.yaml as apiKey: ${E2B_API_KEY} — a named secret you set with parel secrets set (or that parel deploy uploads from your local environment), not a model provider key.
# Beta
PAREL is in beta. Prices are in USD and exclude taxes. We will announce price changes in advance; they apply from your next billing period.
Questions about pricing or volume needs? contact@parel.sh.