Rate Limits and Usage

Learn how execution limits work across plans, when you will see 429 responses, and how to make the most of your quota.

How execution limits work

Every time you run a Flow, it counts as one execution. Each plan includes an execution pool that resets on a daily or monthly schedule.

Burst rate is the maximum number of Flow executions allowed in a 10-second window. If you exceed it, the API returns 429 until the window resets, which can take up to 10 seconds.

On paid plans, your monthly execution pool is a soft cap. You are not blocked when you use it up. Overage is billed at the end of the period. On paid plans, you only see 429 responses when you exceed the burst rate. These limits apply to Flow executions, not to individual AI model calls. Model providers enforce their own limits separately, especially when you use Platform Keys vs. Bring Your Own Key (BYOK).

Slow mode on Build

On the Build plan, your first 50 Flow executions each day run at full speed. After that, your workspace enters slow mode.

• Each request is delayed by 10 seconds
• Up to 10 requests per hour are allowed
• Requests are still processed, so you can keep testing and demoing

If you reach the hourly limit in slow mode, the API returns 429 and includes a Retry-After header to tell you when to try again. Upgrading to a paid plan removes slow mode and gives you a monthly execution pool with a higher burst rate.

You can check whether you are in slow mode from the dashboard header. Look for the Slow mode indicator.

Platform keys and BYOK

Your API key setup affects which AI provider limits apply to your requests.

Platform keys are shared across Runtype users. They are useful for getting started and for development, but they can be throttled during peak usage.

BYOK gives you dedicated limits from your AI provider. Use it for production workloads or when you need more consistent response times. To set up BYOK, follow Connecting AI model providers.

Check your usage

You can monitor usage from the dashboard and from API responses.

• The header shows your current execution usage or a Slow mode indicator
• Use Billing and plans to review plan details and upgrade options

Execution responses also include headers such as X-Quota-Remaining and RateLimit-Remaining. When you have used 80% or more of your execution pool, the response includes X-Quota-Warning.

Tips for staying within your limits

• Cache repeated results in Records when you do not need to run the same Flow again
• Lower concurrency and add delay steps for large workloads
• Add retry logic that respects the Retry-After header on 429 responses

Upgrade when your usage grows

You can upgrade from Settings → Billing or from upgrade prompts shown near your limit.

• Build → Startup: move from 50/day with slow mode to 15,000/month
• Startup → Growth → Team: increase both your execution pool and burst rate
• Team → Enterprise: get custom limits, unlimited executions, and dedicated support

If you are consistently using 80% or more of your pool, it is a good time to upgrade. Higher plans also give you a higher burst rate.

If you are hitting limits in production

1. Check your current usage in Settings → Billing
2. Switch to BYOK if you are using platform keys
3. Add delays between high-frequency steps in your Flows
4. Pause non-critical scheduled jobs temporarily
5. Upgrade your plan if your current tier no longer fits your traffic

For plan details, see Billing and plans. For model configuration, see Connecting AI model providers and Platform Keys vs. Bring Your Own Key (BYOK).

Next steps

• Connecting AI model providers to set up BYOK for dedicated provider limits
• Platform Keys vs. Bring Your Own Key (BYOK) to choose the right model access setup
• Billing and plans to review plan options and upgrade paths
• What are Flows? to understand what counts as an execution
• What are Records? to reduce unnecessary executions by caching reusable results