You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Most platform components are fully serverless and incur zero cost when idle: DynamoDB (PAY_PER_REQUEST), Lambda, API Gateway, ECS Fargate (cluster is free, when enabled), AgentCore Runtime (per-session), Bedrock (per-token), and Cognito (free tier). The always-on cost floor (~$140–150/month) is dominated by VPC networking infrastructure (NAT Gateway + 7 interface endpoints across 2 AZs) which is required for private subnet connectivity to AWS services and GitHub. See the [Deployment guide](../guides/DEPLOYMENT_GUIDE.md) for the full scale-to-zero breakdown.
26
+
Most platform components are fully serverless and incur zero cost when idle: DynamoDB (PAY_PER_REQUEST, 6 tables), Lambda, API Gateway, S3 (trace artifacts auto-expire in 7 days), SQS (fanout DLQ), ECS Fargate (cluster is free, when enabled), AgentCore Runtime (per-session), Bedrock (per-token), and Cognito (free tier). The stranded task reconciler adds <$0.01/month even when idle (288 Lambda invocations/day, early-exit). The always-on cost floor (~$140–150/month) is dominated by VPC networking infrastructure (NAT Gateway + 7 interface endpoints across 2 AZs) which is required for private subnet connectivity to AWS services and GitHub. See the [Deployment guide](../guides/DEPLOYMENT_GUIDE.md) for the full scale-to-zero breakdown.
24
27
25
28
## Per-task variable costs
26
29
@@ -35,10 +38,14 @@ Assuming a typical task: 1–2 hours, Claude Sonnet, ~100K input tokens, ~20K ou
35
38
|**Bedrock tokens (dominant)**| $2–15 | Varies widely by model, task complexity, and turn count. Claude Sonnet: ~$3/M input tokens, ~$15/M output tokens. A 50-turn task with 100K input + 20K output per turn ≈ 5M input + 1M output ≈ $15 + $15 = $30 at list price. Prompt caching reduces this significantly (up to 90% for cache hits). Typical range: $2–15 after caching. |
36
39
| AgentCore Runtime compute | $0.10–0.50 | 2 vCPU / 8 GB for 1–2 hours. Pricing model is per-session based on vCPU-hours and GB-hours. |
| Lambda fanout consumer | <$0.01 | Triggered per batch of task events (batch size 100, 5 s window). Typically 5–20 invocations per task at 256 MB. Negligible. |
42
+
| Lambda nudge / trace / events | <$0.01 | On-demand per user request. Negligible unless heavily polled. |
Most platform components are fully serverless and incur zero cost when idle: DynamoDB (PAY_PER_REQUEST), Lambda, API Gateway, ECS Fargate (cluster is free, when enabled), AgentCore Runtime (per-session), Bedrock (per-token), and Cognito (free tier). The always-on cost floor (~$140–150/month) is dominated by VPC networking infrastructure (NAT Gateway + 7 interface endpoints across 2 AZs) which is required for private subnet connectivity to AWS services and GitHub. See the [Deployment guide](/getting-started/deployment-guide) for the full scale-to-zero breakdown.
30
+
Most platform components are fully serverless and incur zero cost when idle: DynamoDB (PAY_PER_REQUEST, 6 tables), Lambda, API Gateway, S3 (trace artifacts auto-expire in 7 days), SQS (fanout DLQ), ECS Fargate (cluster is free, when enabled), AgentCore Runtime (per-session), Bedrock (per-token), and Cognito (free tier). The stranded task reconciler adds <$0.01/month even when idle (288 Lambda invocations/day, early-exit). The always-on cost floor (~$140–150/month) is dominated by VPC networking infrastructure (NAT Gateway + 7 interface endpoints across 2 AZs) which is required for private subnet connectivity to AWS services and GitHub. See the [Deployment guide](/getting-started/deployment-guide) for the full scale-to-zero breakdown.
28
31
29
32
## Per-task variable costs
30
33
@@ -39,10 +42,14 @@ Assuming a typical task: 1–2 hours, Claude Sonnet, ~100K input tokens, ~20K ou
39
42
|**Bedrock tokens (dominant)**| $2–15 | Varies widely by model, task complexity, and turn count. Claude Sonnet: ~$3/M input tokens, ~$15/M output tokens. A 50-turn task with 100K input + 20K output per turn ≈ 5M input + 1M output ≈ $15 + $15 = $30 at list price. Prompt caching reduces this significantly (up to 90% for cache hits). Typical range: $2–15 after caching. |
40
43
| AgentCore Runtime compute | $0.10–0.50 | 2 vCPU / 8 GB for 1–2 hours. Pricing model is per-session based on vCPU-hours and GB-hours. |
| Lambda fanout consumer | <$0.01 | Triggered per batch of task events (batch size 100, 5 s window). Typically 5–20 invocations per task at 256 MB. Negligible. |
46
+
| Lambda nudge / trace / events | <$0.01 | On-demand per user request. Negligible unless heavily polled. |
0 commit comments