feat: add project aware observability support for harness

README.md

@@ -139,7 +139,11 @@ agentcore                          # interactive TUI
 │   │   ├── runtime                # use the existing Runtime invoke experience
 │   │   └── harness                # use the existing Harness invoke experience
 │   ├── status                     # inspect deployed project resources (TUI when run bare)
-│   └── build                      # synthesize the project's CloudFormation templates
+│   ├── build                      # synthesize the project's CloudFormation templates
+│   ├── log
+│   │   └── runtime                # resolve a project Runtime and inspect its logs
+│   └── traces
+│       └── runtime
 └── config                         # read/write global config values
 ```
 
@@ -186,6 +190,48 @@ agentcore project invoke harness \
 Use `--target` to select a deployment target. When a project declares exactly
 one resource of the requested type, `--name` may be omitted.
 
+### Inspect project logs
+
+Project logging resolves a logical resource name through the selected
+deployment target, so physical IDs and deployment regions do not need to be
+supplied:
+
+```bash
+agentcore project log runtime
+agentcore project log runtime --name checkout --target production
+agentcore project log runtime --name checkout --since 1h --level error
+agentcore project log harness
+agentcore project log harness --name support --target production
+agentcore project log harness --name support --since 1h --level error
+```
+
+When the project declares exactly one resource of the requested type, `--name`
+may be omitted. For Harnesses, the CLI also resolves the managed Harness to its
+underlying Runtime before reading CloudWatch. Use the imperative
+`agentcore runtime logs` or `agentcore harness logs` commands when addressing a
+physical resource directly or working outside a project.
+
+### Inspect project traces
+
+Project tracing uses the same logical resource and deployment target
+resolution, then lists or downloads traces from the resolved Runtime's
+deployment region:
+
+```bash
+agentcore project traces runtime list
+agentcore project traces runtime list --name checkout --target production --since 30m
+agentcore project traces runtime get <traceId> --name checkout --output trace.json
+agentcore project traces harness list
+agentcore project traces harness list --name support --target production --since 30m
+agentcore project traces harness get <traceId> --name support --output trace.json
+```
+
+When the project declares exactly one resource of the requested type, `--name`
+may be omitted. For Harnesses, the CLI resolves the underlying Runtime before
+querying its traces. Use the imperative `agentcore runtime traces` or
+`agentcore harness traces` commands when addressing a physical resource
+directly or working outside a project.
+
 ### Examples
 
 ```bash
@@ -251,7 +297,7 @@ agentcore runtime version list --id <runtimeId> --max-results 20
 agentcore runtime endpoint get --id <runtimeId> --qualifier DEFAULT
 agentcore runtime endpoint list --id <runtimeId> --max-results 20
 
-# Follow a Runtime's logs live (Ctrl+C to stop); inside a project --id is optional
+# Follow a Runtime's logs live by physical ID (Ctrl+C to stop)
 agentcore runtime logs --id <runtimeId>
 agentcore runtime logs --id <runtimeId> --level error --query "database"
 
@@ -263,6 +309,13 @@ agentcore runtime logs --id <runtimeId> --since 2026-08-30T12:00:00Z --until now
 agentcore runtime traces list --id <runtimeId> --since 30m
 agentcore runtime traces get <traceId> --id <runtimeId> --output trace.json
 
+# Resolve project resources by logical name and deployment target
+agentcore project log harness --name support --target production --since 1h
+agentcore project traces runtime list --name checkout --target production --since 30m
+agentcore project traces runtime get <traceId> --name checkout --output trace.json
+agentcore project traces harness list --name support --target production --since 30m
+agentcore project traces harness get <traceId> --name support --output trace.json
+
 # Inspect AgentCore Memories without project configuration or deployment
 agentcore memory get --id <memoryId>
 agentcore memory get --id <memoryId> --view without_decryption

src/core/eval.tsx

@@ -115,8 +115,8 @@ import {
   runtimeLogGroup,
   sanitizeQueryValue,
   type InsightsRowLimit,
-} from "./observability";
-import { CloudWatchClient } from "./observability/index";
+  CloudWatchClient,
+} from "./observability/index";
 import type {
   BatchEvaluationDetail,
   CodeBasedUpdate,

src/core/index.tsx

@@ -8,11 +8,9 @@ import { HarnessClient } from "./harness";
 import { IdentityClient } from "./identity";
 import { MemoryClient } from "./memory";
 import { PolicyClient } from "./policy";
-import { ObservabilityClient } from "./observability";
-import { CloudWatchClient } from "./observability/index";
+import { CloudWatchClient, ObservabilityClient } from "./observability/index";
 import { RuntimeClient } from "./runtime";
 import type { OpenRuntimeShell } from "./runtime";
-import { FsReadWriteJson } from "../io";
 import type {
   AwsClients,
   AwsCredentials,
@@ -113,14 +111,7 @@ export class CoreClient implements AwsClients {
       cloudWatch,
     );
 
-    // Observability resolves a project's deployed runtime from its stack
-    // outputs, so it reads aws-targets.json through the same JSON layer the
-    // project manager uses.
-    this.observability = new ObservabilityClient(cloudWatch, {
-      readJson: new FsReadWriteJson({
-        logger: this.logger.child({ module: "observability" }),
-      }),
-    });
+    this.observability = new ObservabilityClient(cloudWatch);
 
     this.projectManager = new FsProjectManager({
       logger: this.logger.child({ module: "projectManager" }),