View a markdown version of this page

Rate limit API examples - Amazon Bedrock AgentCore

Rate limit API examples

The following examples show how to manage rate limits using the AWS CLI and AWS Python SDK (Boto3).

Create a per-target requests-per-second limit

The following example creates a rate limit that controls requests per second on a per-target basis. A specific target gets 100 RPS, and all other targets each get their own 10 RPS bucket.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control create-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --dimension-keys '["targetName"]' \ --description "Per-target RPS limit" \ --entries '[ { "dimensions": {"targetName": "my-high-traffic-target"}, "requests": [ { "rate": 100, "period": "second" } ] }, { "dimensions": {"targetName": "*"}, "requests": [ { "rate": 10, "period": "second" } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.create_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", dimensionKeys=["targetName"], description="Per-target RPS limit", entries=[ { "dimensions": {"targetName": "my-high-traffic-target"}, "requests": [ { "rate": 100, "period": "second", } ], }, { "dimensions": {"targetName": "*"}, "requests": [ { "rate": 10, "period": "second", } ], }, ], ) print(f"Rate Limit ID: {response['rateLimitId']}")

Create a per-caller requests-per-minute limit

The following example creates a rate limit based on the caller’s JWT sub claim. Premium users get 300 RPM, and all other users get 60 RPM.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control create-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --dimension-keys '["$.context.jwt.sub"]' \ --description "Per-caller RPM by subscription tier" \ --entries '[ { "dimensions": {"$.context.jwt.sub": "premium-user-001"}, "requests": [ { "rate": 300, "period": "minute" } ] }, { "dimensions": {"$.context.jwt.sub": "premium-user-002"}, "requests": [ { "rate": 300, "period": "minute" } ] }, { "dimensions": {"$.context.jwt.sub": "*"}, "requests": [ { "rate": 60, "period": "minute" } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.create_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", dimensionKeys=["$.context.jwt.sub"], description="Per-caller RPM by subscription tier", entries=[ { "dimensions": {"$.context.jwt.sub": "premium-user-001"}, "requests": [ { "rate": 300, "period": "minute", } ], }, { "dimensions": {"$.context.jwt.sub": "premium-user-002"}, "requests": [ { "rate": 300, "period": "minute", } ], }, { "dimensions": {"$.context.jwt.sub": "*"}, "requests": [ { "rate": 60, "period": "minute", } ], }, ], ) print(f"Rate Limit ID: {response['rateLimitId']}")

Create a multi-dimension limit with tokens

The following example creates a rate limit with three dimension keys that limits both requests and tokens per minute. Each caller gets their own budget scoped to a specific target and model.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control create-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --dimension-keys '["targetName", "qualifiedModelId", "$.context.jwt.sub"]' \ --description "Per-caller token and request budget per target and model" \ --entries '[ { "dimensions": {"targetName": "my-inference-target", "qualifiedModelId": "anthropic.claude-3-sonnet-20240229-v1:0", "$.context.jwt.sub": "*"}, "requests": [ { "rate": 100, "period": "minute" } ], "tokens": [ { "rate": 50000, "period": "minute" } ] }, { "dimensions": {"targetName": "*", "qualifiedModelId": "*", "$.context.jwt.sub": "*"}, "requests": [ { "rate": 30, "period": "minute" } ], "tokens": [ { "rate": 10000, "period": "minute" } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.create_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", dimensionKeys=["targetName", "qualifiedModelId", "$.context.jwt.sub"], description="Per-caller token and request budget per target and model", entries=[ { "dimensions": {"targetName": "my-inference-target", "qualifiedModelId": "anthropic.claude-3-sonnet-20240229-v1:0", "$.context.jwt.sub": "*"}, "requests": [ { "rate": 100, "period": "minute", } ], "tokens": [ { "rate": 50000, "period": "minute", } ], }, { "dimensions": {"targetName": "*", "qualifiedModelId": "*", "$.context.jwt.sub": "*"}, "requests": [ { "rate": 30, "period": "minute", } ], "tokens": [ { "rate": 10000, "period": "minute", } ], }, ], ) print(f"Rate Limit ID: {response['rateLimitId']}")

Create a connection rate limit

The following example creates a connection rate limit per model. This limits the number of concurrent in-flight requests to each model, which is useful for protecting targets that have limited concurrency capacity.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control create-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --dimension-keys '["qualifiedModelId"]' \ --description "Connection rate limit per model" \ --entries '[ { "dimensions": {"qualifiedModelId": "anthropic.claude-3-sonnet-20240229-v1:0"}, "connections": [ { "rate": 50, "period": "second" } ] }, { "dimensions": {"qualifiedModelId": "*"}, "connections": [ { "rate": 20, "period": "second" } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.create_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", dimensionKeys=["qualifiedModelId"], description="Connection rate limit per model", entries=[ { "dimensions": {"qualifiedModelId": "anthropic.claude-3-sonnet-20240229-v1:0"}, "connections": [ { "rate": 50, "period": "second", } ], }, { "dimensions": {"qualifiedModelId": "*"}, "connections": [ { "rate": 20, "period": "second", } ], }, ], ) print(f"Rate Limit ID: {response['rateLimitId']}")

Block a caller with rate zero

The following example blocks a specific caller by setting their rate to 0. This denies all requests from the blocked caller once the change propagates (up to 30 seconds).

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control create-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --dimension-keys '["$.context.jwt.sub"]' \ --description "Block abusive caller" \ --entries '[ { "dimensions": {"$.context.jwt.sub": "blocked-user-789"}, "requests": [ { "rate": 0, "period": "second" } ] }, { "dimensions": {"$.context.jwt.sub": "*"}, "requests": [ { "rate": 100, "period": "second" } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.create_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", dimensionKeys=["$.context.jwt.sub"], description="Block abusive caller", entries=[ { "dimensions": {"$.context.jwt.sub": "blocked-user-789"}, "requests": [ { "rate": 0, "period": "second", } ], }, { "dimensions": {"$.context.jwt.sub": "*"}, "requests": [ { "rate": 100, "period": "second", } ], }, ], ) print(f"Rate Limit ID: {response['rateLimitId']}")

Get a rate limit

The following example retrieves the details of a specific rate limit.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control get-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --rate-limit-id rl-abc123def456
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.get_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", rateLimitId="rl-abc123def456", ) print(f"Status: {response['status']}") print(f"Dimension Keys: {response['dimensionKeys']}") print(f"Entries: {len(response['entries'])}")

Update a rate limit

The following example updates the entries of an existing rate limit to increase the rate for a specific target.

Note

The dimensionKeys field is immutable after creation. To change dimension keys, delete the rate limit and create a new one.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control update-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --rate-limit-id rl-abc123def456 \ --description "Updated per-target RPS limit" \ --entries '[ { "dimensions": {"targetName": "my-high-traffic-target"}, "requests": [ { "rate": 200, "period": "second" } ] }, { "dimensions": {"targetName": "*"}, "requests": [ { "rate": 20, "period": "second" } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.update_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", rateLimitId="rl-abc123def456", description="Updated per-target RPS limit", entries=[ { "dimensions": {"targetName": "my-high-traffic-target"}, "requests": [ { "rate": 200, "period": "second", } ], }, { "dimensions": {"targetName": "*"}, "requests": [ { "rate": 20, "period": "second", } ], }, ], ) print(f"Status: {response['status']}")

List rate limits

The following example lists all rate limits for a gateway.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control list-gateway-rate-limits \ --gateway-identifier my-gateway-abc1234567
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") paginator = client.get_paginator("list_gateway_rate_limits") for page in paginator.paginate(gatewayIdentifier="my-gateway-abc1234567"): for rate_limit in page["rateLimits"]: print( f"ID: {rate_limit['rateLimitId']}, " f"Status: {rate_limit['status']}, " f"Dimensions: {rate_limit['dimensionKeys']}" )
Note

Results might be paginated. Use the nextToken value from the response to retrieve additional pages, or use the Boto3 paginator as shown above.

Delete a rate limit

The following example deletes a rate limit from a gateway.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control delete-gateway-rate-limit \ --gateway-identifier my-gateway-abc1234567 \ --rate-limit-id rl-abc123def456
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.delete_gateway_rate_limit( gatewayIdentifier="my-gateway-abc1234567", rateLimitId="rl-abc123def456", ) print(f"Status: {response['status']}")

Batch put rate limits

The following example creates or updates multiple rate limits in a single call. Batch put uses upsert semantics — if a rate limit with the same dimension keys exists, it is updated; otherwise, a new rate limit is created.

Note

Batch put is idempotent. You can safely retry failed calls without creating duplicate rate limits.

Example
AWS CLI
  1. Run the following command:

    aws bedrock-agentcore-control batch-put-gateway-rate-limits \ --gateway-identifier my-gateway-abc1234567 \ --rate-limits '[ { "dimensionKeys": ["targetName"], "description": "Per-target RPS", "entries": [ { "dimensions": {"targetName": "*"}, "requests": [{"rate": 50, "period": "second"}] } ] }, { "dimensionKeys": ["$.context.jwt.sub"], "description": "Per-caller RPM", "entries": [ { "dimensions": {"$.context.jwt.sub": "*"}, "requests": [{"rate": 120, "period": "minute"}] } ] } ]'
AWS Python SDK (Boto3)
  1. import boto3 client = boto3.client("bedrock-agentcore-control", region_name="us-west-2") response = client.batch_put_gateway_rate_limits( gatewayIdentifier="my-gateway-abc1234567", rateLimits=[ { "dimensionKeys": ["targetName"], "description": "Per-target RPS", "entries": [ { "dimensions": {"targetName": "*"}, "requests": [ {"rate": 50, "period": "second"} ], } ], }, { "dimensionKeys": ["$.context.jwt.sub"], "description": "Per-caller RPM", "entries": [ { "dimensions": {"$.context.jwt.sub": "*"}, "requests": [ {"rate": 120, "period": "minute"} ], } ], }, ], ) for result in response["rateLimits"]: print(f"ID: {result['rateLimitId']}, Status: {result['status']}")