Data Streams Rate Limits and Fair Use Policy
To ensure reliable access to Chainlink Data Streams for all consumers, rate limits are enforced on API and WebSocket usage. This page describes what constitutes fair usage.
Fair usage definition
Rate limits are defined by your service agreement. For users with a self-service subscription purchased through app.chain.link, fair usage is defined as:
- 20 concurrent WebSocket connections
- 10 requests per second per endpoint on the REST service
These limits apply per user ID. For example, your account is limited to 10 requests per second on each of the following endpoints (limits are not cumulative across endpoints):
/api/v1/reports/api/v1/reports/latest/api/v1/reports/page/api/v1/reports/bulk
Users with a contracted service agreement have higher limits defined by their agreement. Contact us for higher rate limits.
Requests that exceed the applicable limit receive an HTTP 429 (Too Many Requests) response.
Report retrieval
Reports are pushed as soon as they are available. Sending a higher volume of REST requests to retrieve the latest report will not return it any faster, and adds unnecessary load to the service. Use the WebSocket stream for timely report delivery rather than polling a REST endpoint.
Prohibited: redistribution via proxy
Proxying Chainlink Data Streams with the intent to share or redistribute data to a wider audience (for example, re-serving stream data to third parties who are not party to your license or agreement) is strictly prohibited under the Terms of Service. Each consumer must connect and authenticate directly under their own credentials.
Best practices to avoid rate limiting
- Rely on the WebSocket stream for report delivery rather than polling a REST endpoint in a tight loop.
- Ensure your client properly closes WebSocket connections on disconnect and reconnect to avoid connection leaks.
- Implement exponential backoff on reconnect logic.
- If you require higher throughput or connection limits than the defaults, reach out proactively rather than working around the limit — we're happy to discuss options for your use case.