Cost per model and request, latency and throttling, alerts, and savings recommendations — from the invocation logs Bedrock already writes.
Estimate your monthly spend and see exactly how much Batch API and prompt caching would save you.
BedrockOps reads your model invocation logs through a read-only IAM role — nothing sits in your request path.
What each call costs, by model, region, and account — including prompt-cache reads and writes. Tag calls by team or feature via the SDK or Bedrock request metadata.
Throttled and failed calls per model and region, so you see capacity problems before your users do.
P50, P95, and P99 latency per model, from invocation logs or the SDK.
Cost per hour, spend, latency, throttle rate, and error rate thresholds, delivered by email or Slack.
Prompt caching opportunities, batch inference candidates, cheaper-model suggestions, and a Provisioned Throughput break-even check.
Connect several AWS accounts and regions and see spend in one place; roll up a whole team on the Team plan.
For individual developers
For teams shipping AI products
For organizations at scale
Prices in USD per month, excluding VAT. Upgrade, switch, or cancel anytime from Settings.
Full setup in the docs.
Turn on Bedrock invocation logging, then give BedrockOps a read-only role:
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "ReadBedrockInvocationLogs",
"Effect": "Allow",
"Action": "logs:FilterLogEvents",
"Resource": [
"arn:aws:logs:*:*:log-group:/aws/bedrock/modelinvocations",
"arn:aws:logs:*:*:log-group:/aws/bedrock/modelinvocations:*"
]
}
]
}One file, no dependencies. Tracks every boto3 Bedrock call.
curl -O https://bedrockops.site/sdk/bedrockops.py import bedrockops bo = bedrockops.Client(api_key="bo_...") bo.patch_boto3()
Send Converse calls through BedrockOps; they run with your role.
POST https://api.bedrockops.site
/proxy/converse
X-API-Key: bo_...
{"connection_id": "...",
"model_id": "us.anthropic...",
"body": {"messages": [...]}}