diff --git a/API.md b/API.md
index c791223..87c60bc 100644
--- a/API.md
+++ b/API.md
@@ -1339,12 +1339,16 @@ Submit a blob directly to the batcher queue. Structure and cryptographic proofs
 
 **Rate-limited `429`**
 
+Per client IP, `API_RATE_LIMIT_MAX` requests per minute across every `/v1/*`
+route (default 600; `/v1/health` and `/v1/health/sync` are exempt). The
+`x-ratelimit-limit`, `x-ratelimit-remaining` and `x-ratelimit-reset` headers
+are set on every counted response; back off on `429` instead of retry-looping.
+
 ```json
 {
-  "success": false,
-  "error": "Rate limit exceeded",
-  "message": "Too many requests. Please retry after 60 seconds.",
-  "retryAfter": 60
+  "statusCode": 429,
+  "error": "RATE_LIMITED",
+  "reason": "Too many requests — please wait before retrying."
 }
 ```
 
diff --git a/FRONTEND-API-HANDOFF.md b/FRONTEND-API-HANDOFF.md
index 76c6a97..a8dbbad 100644
--- a/FRONTEND-API-HANDOFF.md
+++ b/FRONTEND-API-HANDOFF.md
@@ -7,7 +7,7 @@ from memory. The backend will be **redeployed from zero** with these changes —
 is no compatibility window and no legacy fallback: the old API simply stops existing.
 
 Base URL: `http://<host>:9999`. All offer-related endpoints moved from `/api/*` to
-`/v1/*`. Rate limit: **60 requests/min per IP** across all routes → HTTP 429
+`/v1/*`. Rate limit: **600 requests/min per IP** (`API_RATE_LIMIT_MAX`) across all routes → HTTP 429
 `{ "error": "RATE_LIMITED", "reason": "..." }`. On 429, back off; do not retry-loop.
 
 ---
diff --git a/README.md b/README.md
index 656b19b..6c87d47 100644
--- a/README.md
+++ b/README.md
@@ -437,6 +437,7 @@ Fund the `celestia1...` address shown by `celestia state account-address` with T
 | `CELESTIA_POLLING_INTERVAL_MS` | optional | Sync cadence. Defaults: devnet 6 000 ms, mainnet 30 000 ms. |
 | `MIDNIGHT_START_BLOCK` | yes | Numeric block height to start Midnight sync from. |
 | `NTP_START_TIME` | optional | NTP reference timestamp; resumed from DB when unset. |
+| `API_RATE_LIMIT_MAX`, `API_RATE_LIMIT_ALLOWLIST` | optional | Per-client-IP budget for `/v1/*` per minute (default 600; `/v1/health*` exempt) and comma-separated IPs exempt from it. Size it for a solver plus its console on one address (~450/min worst case); `deploy/` sets 6000. |
 | `BATCHER_SUBMIT_TIMEOUT_MS` | optional | Absolute batcher fetch + receipt-body deadline; default 310 000 ms, bounded to 1 000–600 000 ms. |
 | `API_SSE_MAX_CONNECTIONS` | optional | Per-node concurrent `/v1/offers/stream` cap; default 100. Excess clients receive `503 SSE_CAPACITY`. |
 | `API_UPDATES_MAX_CONNECTIONS` | optional | Per-node concurrent `/v1/offers/updates` websocket cap; default 100. Excess clients are refused the connection (this endpoint's refusals are disconnects, not HTTP statuses — see API.md). |
diff --git a/deploy/.env.example b/deploy/.env.example
index d73a6c6..4bb4617 100644
--- a/deploy/.env.example
+++ b/deploy/.env.example
@@ -214,7 +214,7 @@ SOLVER_FRONTEND_POLL_MS=
 SOLVER_FRONTEND_HISTORY_LIMIT=
 
 # ── kernel ───────────────────────────────────────────────────────────────────
-# The dev per-IP budget (60/min) throttles a co-located solver during a
+# The code default per-IP budget (600/min) still throttles a co-located solver during a
 # page-through plus settlement polls, and would throttle the E2E driver itself.
 API_RATE_LIMIT_MAX=6000
 API_RATE_LIMIT_ALLOWLIST=
diff --git a/packages/node/api.ts b/packages/node/api.ts
index 28c98ec..6f05295 100644
--- a/packages/node/api.ts
+++ b/packages/node/api.ts
@@ -101,7 +101,7 @@ export const apiRouter: StartConfigApiRouter = async function (
   server: any,
   dbConn: any,
 ): Promise<void> {
-  // Per-IP request budget (default 60/min) — applied to every route in this
+  // Per-IP request budget (default 600/min) — applied to every route in this
   // router.
   //
   // `statusCode` is load-bearing, not decoration: @fastify/rate-limit THROWS
@@ -745,7 +745,7 @@ export const apiRouter: StartConfigApiRouter = async function (
   // Uses effectstream.effectstream_blocks for NTP and
   // effectstream.sync_protocol_pagination for parallel chains.
   // Chain tips are fetched from the Midnight indexer / Celestia RPC and cached 60 s.
-  // Exempt from the 60/min API budget — UIs poll this as a liveness probe.
+  // Exempt from the per-IP API budget — UIs poll this as a liveness probe.
   server.get("/v1/health/sync", { config: { rateLimit: false } }, async () => {
     return getSyncStatus(dbConn);
   });
diff --git a/packages/node/env.ts b/packages/node/env.ts
index 76e5de3..5099305 100644
--- a/packages/node/env.ts
+++ b/packages/node/env.ts
@@ -129,7 +129,7 @@ export const OFFER_MAX_BYTES = parseInt(
 // building a router gets them — the same reason isTokenRegistryEnabled below
 // is a function.
 export const apiRateLimitMax = (): number =>
-  parseInt(getEnv("API_RATE_LIMIT_MAX") ?? "60");
+  parseInt(getEnv("API_RATE_LIMIT_MAX") ?? "600");
 
 export const apiRateLimitAllowList = (): string[] =>
   (getEnv("API_RATE_LIMIT_ALLOWLIST") ?? "")