A value used in filter comparisons, supporting different data types.

A content block for ground truth data in evaluation reference inputs. Supports text content for expected responses and assertions.

The input for a ground truth conversation turn.

Email address with validation for standard email format

A filter for narrowing down agent traces from CloudWatch Logs based on key-value comparisons.

The tool description content.

Ground truth data for a single conversation turn.

A content block within a tool result.

JWT Key ID (kid) from the JWT header Identifies which key was used to sign the JWT Must be alphanumeric with hyphens and underscores, 1-255 characters

OAuth2 authentication information for third-party providers. Supports Google, Apple, X, Telegram, and GitHub providers.

Phone number in E.164 format (e.g., +1234567890)

Maps a tool name to its JSON path within a configuration bundle.

A tool description input containing the tool name and its current description.

Configuration for HTTP Basic Authentication using credentials stored in Amazon Web Services Secrets Manager. The secret must contain a JSON object with username and password string fields. Username allows alphanumeric characters and @._+=- symbols (pattern: ^[a-zA-Z0-9@._+=\-]+$). Password allows alphanumeric characters and @._+=-!#$%&* symbols (pattern: ^[a-zA-Z0-9@._+=\-!#$%&*]+$). Both fields have a maximum length of 256 characters.

The expected tool call trajectory for trajectory-based evaluation.

Contains the content of a memory item.

Aggregated statistics for an evaluator.

A block of reasoning text from the model.

Configuration for an OAuth 2.0 credential provider used to authenticate tool calls.

The key of an HTTP header.

The value of an HTTP header.

Authentication method using JWT with key ID and subject claims.

LinkedAccount using an email address.

Authentication method using OAuth2 providers. Supports Google, Apple, X, Telegram, and GitHub providers.

LinkedAccount using a phone number in E.164 format.

A confidence interval for a statistical measurement.

A rule configuration for filtering agent traces from CloudWatch Logs.

A reference to an evaluator used for recommendation assessment.

Union type representing different proxy authentication methods. Currently supports HTTP Basic Authentication (username and password).

Inline ground truth data containing assertions, expected trajectories, and per-turn expected responses.

The agent card definition for A2A descriptors, including the schema version and inline content that describes the agent's capabilities.

The structured skill definition with a schema version and inline content.

The skill markdown definition for agent skills descriptors.

The MCP server definition with a schema version and inline content. The schemaVersion identifies the version of the MCP server configuration schema.

The MCP tools definition with a protocol version and inline content. The protocolVersion identifies the MCP protocol version that the tools conform to. This differs from schemaVersion in the server definition, which identifies the server configuration schema format.

The value of a memory record metadata entry.

Metadata information associated with this message.

Contains conversational content for an event payload.

Value associated with the eventMetadata key.

Summary statistics for a single evaluator within a batch evaluation.

A delta update to a tool result content block.

Reasoning content from the model.

The result of a tool execution.

A tool use request from the model.

Authentication method for calling a Gateway.

Contains information about resource content.

Represents different linkedAccounts that can be linked to an embedded wallet. This union supports multiple linkedAccount approaches: email, SMS, JWT, and OAuth2.

Statistical results for a treatment variant compared against the control.

A reference to a specific version of a configuration bundle.

A reference to a gateway target.

Configuration for reading agent traces from CloudWatch Logs for recommendation analysis.

A system prompt sourced from a configuration bundle version.

Tool descriptions sourced from a configuration bundle version.

Inline tool description input containing a list of tools.

The Amazon Web Services Secrets Manager location configuration.

The Amazon S3 location configuration of a resource.

Configuration for a customer-managed external proxy server. Includes server location, optional domain-based routing patterns, and authentication credentials.

A time range filter for selecting sessions. Specifies the start and end times to narrow down which sessions are included.

Where to pull ground truth from

The A2A (Agent-to-Agent) descriptor configuration for a registry record.

The agent skills descriptor configuration for a registry record.

A custom descriptor configuration for a registry record.

The MCP (Model Context Protocol) descriptor configuration for a registry record. Contains the server definition and tools definition.

The left-hand side of a memory record metadata filter expression.

The right-hand side of a memory record metadata filter expression.

Contains the payload content for an event.

Left expression of the event metadata filter.

Right expression of the eventMetadatafilter.

An evaluator to run against sessions

A delta update to a reasoning content block.

Delta payload for tool use input.

Start payload for a tool result content block.

Start payload for a tool use content block.

Stores information about a field passed inside a request that resulted in an exception.

A content block within a message.

Configuration for AgentCore Browser.

Configuration for AgentCore Code Interpreter.

Configuration for AgentCore Gateway.

Configuration for an inline function tool. When the agent calls this tool, the tool call is returned to the caller for external execution.

Configuration for connecting to a remote MCP server.

A block of content in a response.

The output for a single tool description recommendation.

Statistics for the control variant in an A/B test.

The contextual information that uniquely identifies a span within the distributed tracing system. Contains session, trace, and span identifiers used to correlate evaluation results with specific agent execution points.

An online evaluation configuration associated with a specific A/B test variant.

The configuration for an A/B test variant.

The configuration specifying where to read agent traces from for recommendation analysis.

The evaluation configuration for assessing recommendation quality.

The system prompt input, either as inline text or from a configuration bundle.

The source of tool descriptions, either inline text or from a configuration bundle.

The location from which to retrieve a certificate.

The location of the browser extension.

Union type representing different proxy configurations. Currently supports external customer-managed proxies.

Filter configuration for narrowing down CloudWatch Logs sessions for evaluation.

Metadata for a specific session in a batch evaluation, including ground truth data and test scenario identifiers.

Contains the descriptor configuration for a registry record. Only the field that matches the record's descriptorType is populated.

Contains the content of a memory record.

Filters to apply to metadata associated with a memory. Specify the metadata key and value in the left and right fields and use the operator field to define the relationship to match.

Custom document type for payment payloads

The list of messages that compose this extraction job.

Contains information about a branch in an AgentCore Memory resource. Branches allow for organizing events into different conversation threads or paths.

Filter expression for retrieving events based on metadata associated with an event.

ARN for a batch evaluation

Unique identifier for a batch evaluation (name-prefixed format)

Human-readable name for a batch evaluation

Aggregated results from a batch evaluation, including session completion counts and evaluator score summaries.

A delta update to a content block.

The start payload for a content block.

Latency metrics for the invocation.

Token usage counts for the invocation.

Configuration union for different tool types.

Contains structured content from a tool result.

A block of input content.

An event that contains incremental output from a command execution. This event streams standard output and standard error content as it becomes available during command execution.

An event that signals the start of content streaming from a command execution. This event is sent when the command begins producing output.

An event that signals the completion of a command execution. This event contains the final status and exit code of the executed command.

A configuration bundle reference in a recommendation result.

Money amount with currency

Embedded Crypto wallet instrument details

Statistical metrics for a single evaluator comparing control and treatment variants.

The contextual information associated with an evaluation, including span context details that identify the specific traces and sessions being evaluated within the agent's execution flow.

The token consumption statistics for language model operations during evaluation. Provides detailed breakdown of input, output, and total tokens used for cost tracking and performance monitoring.

The configuration for a stream that enables programmatic control of a browser session in Amazon Bedrock AgentCore. This stream provides a bidirectional communication channel for sending commands to the browser and receiving responses, allowing agents to automate web interactions such as navigation, form filling, and element clicking.

The configuration for a stream that provides a visual representation of a browser session in Amazon Bedrock AgentCore. This stream enables agents to observe the current state of the browser, including rendered web pages, visual elements, and the results of interactions.

Contains information about an update to an automation stream.

A variant in an A/B test, representing either the control (C) or treatment (T1) configuration.

Configuration for generating system prompt optimization recommendations.

Configuration for generating tool description optimization recommendations.

A certificate to install in the browser or code interpreter session.

Browser enterprise policy configuration.

Browser extension configuration.

Configuration for domains that should bypass all proxies and connect directly to the internet. These bypass rules take precedence over all proxy routing rules.

CloudWatch Logs destination for batch evaluation results

The configuration for reading agent traces from CloudWatch Logs.

Summary information about a registry record.

Contains summary information about a memory record.

X402 payment requirement output

X402 payment requirement input

Contains summary information about a session in an AgentCore Memory resource.

Summary information about a recommendation.

Summary of a payment session for list operations

Summary of a payment instrument for list operations

Metadata information associated with this extraction job.

Contains information about an event in an AgentCore Memory resource.

Contains filter criteria for branches when listing events.

A condensed representation of a code interpreter session in Amazon Bedrock AgentCore. This structure contains key information about a code interpreter session, including identifiers, status, and timestamps, without the full details of the session configuration.

A condensed representation of a browser session in Amazon Bedrock AgentCore. This structure contains key information about a browser session, including identifiers, status, and timestamps, without the full details of the session configuration and streams.

Summary representation for list responses

Contains summary information about an actor in an AgentCore Memory resource.

Summary information about an A/B test.

Event containing a delta update to a content block.

Event indicating the start of a content block.

Event indicating the end of a content block.

Event indicating the start of a message.

Event indicating the end of a message.

Token usage and latency metrics for the invocation.

The exception that occurs when the service encounters an unexpected internal error. This is a temporary condition that will resolve itself with retries. We recommend implementing exponential backoff retry logic in your application.

The exception that occurs when there is an error in the runtime client. This can happen due to network issues, invalid configuration, or other client-side problems. Check the error message for specific details about the error.

The exception that occurs when the input fails to satisfy the constraints specified by the service. Check the error message for details about which input parameter is invalid and correct your request.

A message in the conversation.

Configuration for an Amazon Bedrock model provider.

Configuration for a Google Gemini model provider. Requires an API key stored in AgentCore Identity.

Configuration for an OpenAI model provider. Requires an API key stored in AgentCore Identity.

A skill available to the agent.

A content block in the system prompt.

A tool available to the agent loop.

The exception that occurs when you do not have sufficient permissions to perform an action. Verify that your IAM policy includes the necessary permissions for the operation you are trying to perform.

The output produced by executing code in a code interpreter session in Amazon Bedrock AgentCore. This structure contains the results of code execution, including textual output, structured data, and error information. Agents use these results to generate responses that incorporate computation, data analysis, and visualization.

The exception that occurs when the request conflicts with the current state of the resource. This can happen when trying to modify a resource that is currently being modified by another request, or when trying to create a resource that already exists.

The exception that occurs when the specified resource does not exist. This can happen when using an invalid identifier or when trying to access a resource that has been deleted.

The exception that occurs when the request would cause a service quota to be exceeded. Review your service quotas and either reduce your request rate or request a quota increase.

The exception that occurs when the request was denied due to request throttling. This happens when you exceed the allowed request rate for an operation. Reduce the frequency of requests or implement exponential backoff retry logic in your application.

The result of a key press action.

The result of a key shortcut action.

The result of a key type action.

The result of a mouse click action.

The result of a mouse drag action.

The result of a mouse move action.

The result of a mouse scroll action.

The result of a screenshot action.

Arguments for a key press action.

Arguments for a key shortcut action.

Arguments for a key type action.

Arguments for a mouse click action.

Arguments for a mouse drag action.

Arguments for a mouse move action.

Arguments for a mouse scroll action.

Arguments for a screenshot action.

A structure representing a response chunk that contains exactly one of the possible event types: contentStart, contentDelta, or contentStop.

Coinbase CDP token response

StripePrivy token response containing appId, basicAuthToken, and optionally authorizationSignature

Coinbase CDP token request parameters

StripePrivy token request parameters

The result of a system prompt recommendation, containing the optimized prompt.

The result of a tool description recommendation, containing optimized descriptions.

Available session limits. Currently only budget is supported.

Session spending limits

Details specific to the instrument type

The comprehensive result of an evaluation containing the score, explanation, evaluator metadata, and execution details. Provides both quantitative ratings and qualitative insights about agent performance.

A reference input containing ground truth data for evaluation, scoped to a specific context level (session or trace) through its span context.

Output information returned after processing a memory record operation.

Input structure to update an existing memory record.

Input structure to delete an existing memory record.

Input structure to create a new memory record.

The collection of streams associated with a browser session in Amazon Bedrock AgentCore. These streams provide different ways to interact with and observe the browser session, including programmatic control and visual representation of the browser content.

Contains information about an update to a stream.

This exception is thrown when the JWT bearer token is invalid or not found for OAuth bearer token based access

The evaluation configuration for an A/B test, specifying which online evaluation configurations to use for measuring variant performance.

A filter to restrict which gateway target paths are included in the A/B test.

The configuration for a recommendation, varying by recommendation type.

The service encountered an internal error. Try your request again later.

The request was denied due to request throttling. Reduce the frequency of requests and try again.

Represents the metadata of a memory extraction job such as the message identifiers that compose this job.

The configuration for a browser profile in Amazon Bedrock AgentCore. A browser profile contains persistent browser data such as cookies and local storage that can be saved from one browser session and reused in subsequent sessions. Browser profiles enable continuity for tasks that require authentication, maintain user preferences, or depend on previously stored browser state.

Configuration for routing browser traffic through customer-managed proxy servers. Supports 1-5 proxy servers for domain-based routing and proxy bypass rules.

The configuration that defines the dimensions of a browser viewport in a browser session. The viewport determines the visible area of web content and affects how web pages are rendered and displayed. Proper viewport configuration ensures that web content is displayed correctly for the agent's browsing tasks.

Output destination configuration

Where to pull session spans from

Metadata for the evaluation, including session-specific ground truth data.

A JSON document that represents a structured metadata filter expression. Supports field-level operators ($eq, $ne, $in) and logical operators ($and, $or) on filterable fields (name, descriptorType, version).

The input fails to satisfy the constraints specified by AgentCore. Check your input values and try again.

Contains search criteria for retrieving memory records.

Payment response payload union. We will support versioning by introducing other model in this union

The agent name associated with the payment request. Optional human-readable name for the agent making the API call, used for observability.

Payment request details union

Contains filter criteria for listing sessions.

Filters for querying memory extraction jobs based on various criteria.

Contains filter criteria for listing events.

The streaming events returned by a harness invocation.

Specification of which model to use.

Contains output from a code interpreter stream.

The collection of arguments that specify the operation to perform and its parameters when invoking a tool in Amazon Bedrock AgentCore. Different tools require different arguments, and this structure provides a flexible way to pass the appropriate arguments to each tool type.

The result of a browser action execution. Exactly one member is set, matching the action that was performed.

The browser action to perform. Exactly one member must be set per request.

The streaming output union for the InvokeAgentRuntimeCommand operation. This union delivers typed events: contentStart (first), contentDelta (middle), and contentStop (last).

The request body structure for the InvokeAgentRuntimeCommand operation, containing the command to execute and optional configuration parameters.

VENDOR-SPECIFIC TOKEN RESPONSE CONFIGURATION - Output

VENDOR-SPECIFIC TOKEN REQUEST CONFIGURATION - Input

The result of a recommendation, containing the optimized output.

Payment manager session

Represents a payment instrument

A single token balance entry

Contains information about a memory record in an AgentCore Memory resource.

The statistical results of an A/B test.

An exception thrown when attempting to create a resource with an identifier that already exists.

The input data structure containing agent session spans in OpenTelemetry format. Supports traces from frameworks like Strands (AgentCore Runtime) and LangGraph with OpenInference instrumentation for comprehensive evaluation.

The specification of which trace or span IDs to evaluate within the provided input data. Allows precise targeting of evaluation at different levels: tool calls, traces, or sessions.

The exception that occurs when there is a retryable conflict performing an operation. This is a temporary condition that may resolve itself with retries. We recommend implementing exponential backoff retry logic in your application.

The OAuth2.0 token or user ID that was used to generate the workload access token used for initiating the user authorization flow to retrieve OAuth2.0 tokens.

Updates a browser stream. To use this operation, you must have permissions to perform the bedrock:UpdateBrowserStream action.

Updates a browser stream. To use this operation, you must have permissions to perform the bedrock:UpdateBrowserStream action.

Updates an A/B test's configuration, including variants, traffic allocation, evaluation settings, or execution status.

Updates an A/B test's configuration, including variants, traffic allocation, evaluation settings, or execution status.

Stops a session that is running in an running AgentCore Runtime agent.

Stops a session that is running in an running AgentCore Runtime agent.

Terminates an active code interpreter session in Amazon Bedrock AgentCore. This operation stops the session, releases associated resources, and makes the session unavailable for further use. To stop a code interpreter session, you must specify both the code interpreter identifier and the session ID. Once stopped, a session cannot be restarted; you must create a new session using StartCodeInterpreterSession. The following operations are related to StopCodeInterpreterSession: StartCodeInterpreterSession GetCodeInterpreterSession

Terminates an active code interpreter session in Amazon Bedrock AgentCore. This operation stops the session, releases associated resources, and makes the session unavailable for further use. To stop a code interpreter session, you must specify both the code interpreter identifier and the session ID. Once stopped, a session cannot be restarted; you must create a new session using StartCodeInterpreterSession. The following operations are related to StopCodeInterpreterSession: StartCodeInterpreterSession GetCodeInterpreterSession

Terminates an active browser session in Amazon Bedrock AgentCore. This operation stops the session, releases associated resources, and makes the session unavailable for further use. To stop a browser session, you must specify both the browser identifier and the session ID. Once stopped, a session cannot be restarted; you must create a new session using StartBrowserSession. The following operations are related to StopBrowserSession: StartBrowserSession GetBrowserSession

Terminates an active browser session in Amazon Bedrock AgentCore. This operation stops the session, releases associated resources, and makes the session unavailable for further use. To stop a browser session, you must specify both the browser identifier and the session ID. Once stopped, a session cannot be restarted; you must create a new session using StartBrowserSession. The following operations are related to StopBrowserSession: StartBrowserSession GetBrowserSession

Stops a running batch evaluation. Sessions that have already been evaluated retain their results.

Stops a running batch evaluation. Sessions that have already been evaluated retain their results.

Starts a recommendation job that analyzes agent traces and generates optimization suggestions for system prompts or tool descriptions to improve agent performance.

Starts a recommendation job that analyzes agent traces and generates optimization suggestions for system prompts or tool descriptions to improve agent performance.

Starts a memory extraction job that processes events that failed extraction previously in an AgentCore Memory resource and produces structured memory records. When earlier extraction attempts have left events unprocessed, this job will pick up and extract those as well. To use this operation, you must have the bedrock-agentcore:StartMemoryExtractionJob permission.

Starts a memory extraction job that processes events that failed extraction previously in an AgentCore Memory resource and produces structured memory records. When earlier extraction attempts have left events unprocessed, this job will pick up and extract those as well. To use this operation, you must have the bedrock-agentcore:StartMemoryExtractionJob permission.

Creates and initializes a code interpreter session in Amazon Bedrock AgentCore. The session enables agents to execute code as part of their response generation, supporting programming languages such as Python for data analysis, visualization, and computation tasks. To create a session, you must specify a code interpreter identifier and a name. The session remains active until it times out or you explicitly stop it using the StopCodeInterpreterSession operation. The following operations are related to StartCodeInterpreterSession: InvokeCodeInterpreter GetCodeInterpreterSession StopCodeInterpreterSession

Creates and initializes a code interpreter session in Amazon Bedrock AgentCore. The session enables agents to execute code as part of their response generation, supporting programming languages such as Python for data analysis, visualization, and computation tasks. To create a session, you must specify a code interpreter identifier and a name. The session remains active until it times out or you explicitly stop it using the StopCodeInterpreterSession operation. The following operations are related to StartCodeInterpreterSession: InvokeCodeInterpreter GetCodeInterpreterSession StopCodeInterpreterSession

Creates and initializes a browser session in Amazon Bedrock AgentCore. The session enables agents to navigate and interact with web content, extract information from websites, and perform web-based tasks as part of their response generation. To create a session, you must specify a browser identifier and a name. You can also configure the viewport dimensions to control the visible area of web content. The session remains active until it times out or you explicitly stop it using the StopBrowserSession operation. The following operations are related to StartBrowserSession: GetBrowserSession UpdateBrowserStream SaveBrowserSessionProfile StopBrowserSession InvokeBrowser

Creates and initializes a browser session in Amazon Bedrock AgentCore. The session enables agents to navigate and interact with web content, extract information from websites, and perform web-based tasks as part of their response generation. To create a session, you must specify a browser identifier and a name. You can also configure the viewport dimensions to control the visible area of web content. The session remains active until it times out or you explicitly stop it using the StopBrowserSession operation. The following operations are related to StartBrowserSession: GetBrowserSession UpdateBrowserStream SaveBrowserSessionProfile StopBrowserSession InvokeBrowser

Starts a batch evaluation job that evaluates agent performance across multiple sessions. Batch evaluations pull agent traces from CloudWatch Logs or an existing online evaluation configuration and run specified evaluators and insights against them.

Starts a batch evaluation job that evaluates agent performance across multiple sessions. Batch evaluations pull agent traces from CloudWatch Logs or an existing online evaluation configuration and run specified evaluators and insights against them.

Searches for registry records using semantic, lexical, or hybrid queries. Returns metadata for matching records ordered by relevance within the specified registry.

Searches for registry records using semantic, lexical, or hybrid queries. Returns metadata for matching records ordered by relevance within the specified registry.

Saves the current state of a browser session as a reusable profile in Amazon Bedrock AgentCore. A browser profile captures persistent browser data such as cookies and local storage from an active session, enabling you to reuse this data in future browser sessions. To save a browser session profile, you must specify the profile identifier, browser identifier, and session ID. The session must be active when saving the profile. Once saved, the profile can be used with the StartBrowserSession operation to initialize new sessions with the stored browser state. Browser profiles are useful for scenarios that require persistent authentication, maintaining user preferences across sessions, or continuing tasks that depend on previously stored browser data. The following operations are related to SaveBrowserSessionProfile: StartBrowserSession GetBrowserSession

Saves the current state of a browser session as a reusable profile in Amazon Bedrock AgentCore. A browser profile captures persistent browser data such as cookies and local storage from an active session, enabling you to reuse this data in future browser sessions. To save a browser session profile, you must specify the profile identifier, browser identifier, and session ID. The session must be active when saving the profile. Once saved, the profile can be used with the StartBrowserSession operation to initialize new sessions with the stored browser state. Browser profiles are useful for scenarios that require persistent authentication, maintaining user preferences across sessions, or continuing tasks that depend on previously stored browser data. The following operations are related to SaveBrowserSessionProfile: StartBrowserSession GetBrowserSession

Searches for and retrieves memory records from an AgentCore Memory resource based on specified search criteria. We recommend using pagination to ensure that the operation returns quickly and successfully. To use this operation, you must have the bedrock-agentcore:RetrieveMemoryRecords permission.

Searches for and retrieves memory records from an AgentCore Memory resource based on specified search criteria. We recommend using pagination to ensure that the operation returns quickly and successfully. To use this operation, you must have the bedrock-agentcore:RetrieveMemoryRecords permission.

Response structure for processing a payment

Request structure for processing a payment

Lists sessions in an AgentCore Memory resource based on specified criteria. We recommend using pagination to ensure that the operation returns quickly and successfully. Empty sessions are automatically deleted after one day. To use this operation, you must have the bedrock-agentcore:ListSessions permission.

Lists sessions in an AgentCore Memory resource based on specified criteria. We recommend using pagination to ensure that the operation returns quickly and successfully. Empty sessions are automatically deleted after one day. To use this operation, you must have the bedrock-agentcore:ListSessions permission.

Lists all recommendations in the account, with optional filtering by status.

Lists all recommendations in the account, with optional filtering by status.

Response structure for listing payment sessions

Request structure for listing payment sessions

Response structure for listing payment instruments

Request structure for listing payment instruments

Lists memory records in an AgentCore Memory resource based on specified criteria. We recommend using pagination to ensure that the operation returns quickly and successfully. To use this operation, you must have the bedrock-agentcore:ListMemoryRecords permission.