> For the complete documentation index, see [llms.txt](https://docs.megaeth.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.megaeth.com/developer-docs/rpc/operations-and-limits.md). # Operations and limits This page describes limits applied by the public MegaETH gateway. Method pages define the underlying request and response contracts. ## HTTP request and response limits | Limit | Value | | -------------------------------------- | ------------ | | Default request body | 128 KiB | | Transaction-submission request body | 2.5 MiB | | Simulation and large-read request body | 1.5 MiB | | Batch size | 100 requests | | Batch subrequest budget | 950 | | Response size | 50 MiB | The 2.5 MiB body limit applies to `eth_sendRawTransaction`, `eth_sendRawTransactionSync`, and `realtime_sendRawTransaction`. The 1.5 MiB body limit applies to `eth_call`, `eth_callMany`, `eth_createAccessList`, and `eth_estimateGas`. A body that exceeds its limit is rejected with HTTP `413` and JSON-RPC code `-32099`. The batch subrequest budget counts expanded work rather than only top-level batch items. For example, the calls inside `eth_callMany` contribute to that budget. ## Method-specific limits | Method | Public gateway behavior | | ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------------- | | `eth_call` | Compute gas is capped at 60,000,000. | | `eth_callMany` | At most 100 bundles and 100 total calls, 60,000,000 compute gas per call, and a 25-second timeout. | | `eth_estimateGas` | The node applies a CPU-time limit whose current default is 0.5 seconds. | | `eth_feeHistory` | `blockCount` is capped at 256. | | `eth_getLogs` | The public indexed path does not impose a gateway block-range cap, but backend row, execution-time, memory, and response-size limits still apply. | Use bounded ranges and pagination for log scans and historical backfills. Do not assume that removing a block-range cap makes an unbounded query safe. ## Read rate limits Read requests are limited per client IP in fixed 10-second windows. | Category | Requests per 10 seconds | Typical methods | | -------- | ----------------------- | --------------------------------------------------------------------------------------------------------------------------- | | Instant | 2,000 | `eth_chainId`, `eth_blockNumber`, `net_version`, `eth_accounts`, `web3_clientVersion`, `eth_getBalance`, `eth_getStorageAt` | | Simple | 500 | Basic block and transaction reads not assigned to another category | | Compute | 200 | `eth_call`, `eth_callMany`, `eth_estimateGas`, `eth_createAccessList`, and debug trace methods | | IO-heavy | 200 | `eth_getLogs` and `eth_getBlockReceipts` | Transaction-submission methods are exempt from these read limits. `eth_callMany` is charged by its inner calls rather than as one unit. A rate-limited request returns HTTP `429` with JSON-RPC code `-32005`. Use exponential backoff with jitter for retryable throttling. Reduce concurrency and query size before increasing retry frequency. ## Gateway caching The public gateway has several internal caches. Their eligibility and lifetime depend on the method and selector: immutable block numbers and hashes can use longer-lived entries, while head-following reads use shorter policies or bypass a cache layer. See each method's reference page for its method-specific behavior. The gateway's Workers read-cache layer is limited to: * `eth_getBlockByNumber` * `eth_getBlockReceipts` * `eth_getHeaderByNumber` * `web3_clientVersion` For the block-selecting methods in this layer, only an explicit historical block number, a block hash, or `earliest` is eligible. The `latest`, `pending`, `safe`, and `finalized` tags bypass this layer. Other gateway cache layers can still apply method-specific short-lived policies to head-following reads. Rate-limit accounting differs by request shape: * A single read is rate-limited before cache lookup, so a cache hit consumes its category budget. * A read batch checks the cache first and charges only cache misses against the read-rate budget. Every public response includes `Cache-Control: no-store` for downstream caches. This header does not disable the gateway's internal cache. For methods eligible for the Workers read-cache layer, `X-Workers-Cache-Status` indicates whether that layer returned the response. The gateway does not expose a request header or parameter that bypasses its internal cache. ## WebSocket limits | Limit | Value | | ---------------------------------- | ------------------------------------ | | Connections per IP | 5 | | Subscriptions per connection | 5 | | Client message rate | 5 messages per second per connection | | Idle timeout | 60 seconds | | Maximum message size | 64 KiB | | Addresses in a `logs` filter | 20 | | Topic positions in a `logs` filter | 4 | Use `wss://mainnet.megaeth.com/ws` for Mainnet and `wss://carrot.megaeth.com/ws` for Testnet. Send a lightweight request such as `eth_chainId` at least every 30 seconds to prevent idle disconnection. Reconnect, recreate subscriptions, and reconcile any missed data after a disconnect because notifications are not replayed automatically. The public WebSocket endpoint accepts: * `eth_subscribe` * `eth_unsubscribe` * `eth_sendRawTransaction` * `eth_sendRawTransactionSync` * `realtime_sendRawTransaction` * `eth_chainId` ## Related pages * [Error reference](/developer-docs/rpc/error-codes.md) * [`eth_subscribe`](/developer-docs/rpc/reference/eth_subscribe.md) * [`eth_getLogs`](/developer-docs/rpc/reference/eth_getlogs.md) --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.megaeth.com/developer-docs/rpc/operations-and-limits.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.