Salesforce Docs Portal

Configuring AgentCore Operations

Configure the Amazon Bedrock AgentCore operations. These operations fall into two groups:

  • Memory operations work with memory records in an AgentCore Memory resource: AgentCore - Batch Create Memory Records, AgentCore - Batch Update Memory Records, AgentCore - Batch Delete Memory Records, AgentCore - Get Memory Record, AgentCore - List Memory Records, and AgentCore - Retrieve Memory Records.

  • Runtime and identity operations invoke a deployed agent runtime and obtain AgentCore Identity credentials: AgentCore - Invoke Agent Runtime (Streaming), AgentCore - Invoke Agent Runtime As Bytes (Non Streaming), AgentCore - Stop Runtime Session, AgentCore - Get Workload Access Token, and AgentCore - Get Resource API Key.

Configure the AgentCore - Batch Create Memory Records Operation

The AgentCore - Batch Create Memory Records operation creates multiple memory records in an AgentCore Memory resource in a single batch call.

To configure the AgentCore - Batch Create Memory Records operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Memory Id

      The identifier of the AgentCore Memory resource in which to create the memory records.

    • Create Records

      The list of memory records to create. Each record’s content, namespaces, timestamp, and request identifier are required by AWS; the memory strategy ID and metadata are optional. The timestamp accepts epoch seconds (for example, 1751500000) or an ISO-8601 instant (for example, 2025-07-02T23:46:40Z).

    • Client Token

      An optional idempotency token that AWS uses to deduplicate retried batch requests.

This is the XML for this operation:

<ms-bedrock:agentcore-batch-create-memory-records
  doc:name="AgentCore batch create memory records"
  config-ref="AWS"
  memoryId="#[payload.memoryId]">
  <ms-bedrock:create-records>
    <ms-bedrock:memory-record-create-input-parameters timestamp="1751500000" requestIdentifier="#[uuid()]">
      <ms-bedrock:content text="#[payload.text]"/>
      <ms-bedrock:namespaces>
        <ms-bedrock:namespace value="support/session-123"/>
      </ms-bedrock:namespaces>
    </ms-bedrock:memory-record-create-input-parameters>
  </ms-bedrock:create-records>
</ms-bedrock:agentcore-batch-create-memory-records>

Output Configuration

This operation responds with a JSON payload containing the successfulRecords and failedRecords arrays.

Configure the AgentCore - Batch Update Memory Records Operation

The AgentCore - Batch Update Memory Records operation updates multiple memory records in an AgentCore Memory resource in a single batch call.

To configure the AgentCore - Batch Update Memory Records operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Memory Id

      The identifier of the AgentCore Memory resource whose records are updated.

    • Update Records

      The list of memory records to update. Each record’s memory record ID and timestamp are required by AWS; content, namespaces, memory strategy ID, and metadata are optional. For namespaces, an omitted (null) value leaves the record’s namespaces unchanged, while an explicitly empty list clears them.

This is the XML for this operation:

<ms-bedrock:agentcore-batch-update-memory-records
  doc:name="AgentCore batch update memory records"
  config-ref="AWS"
  memoryId="#[payload.memoryId]">
  <ms-bedrock:update-records>
    <ms-bedrock:memory-record-update-input-parameters memoryRecordId="#[payload.recordId]" timestamp="1751500000">
      <ms-bedrock:content text="#[payload.updatedText]"/>
    </ms-bedrock:memory-record-update-input-parameters>
  </ms-bedrock:update-records>
</ms-bedrock:agentcore-batch-update-memory-records>

Output Configuration

This operation responds with a JSON payload containing the successfulRecords and failedRecords arrays.

Configure the AgentCore - Batch Delete Memory Records Operation

The AgentCore - Batch Delete Memory Records operation deletes multiple memory records from an AgentCore Memory resource in a single batch call.

To configure the AgentCore - Batch Delete Memory Records operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Memory Id

      The identifier of the AgentCore Memory resource whose records are deleted.

    • Delete Records

      The list of memory records to delete. Each record’s memory record ID is required by AWS; the service validates it up front and rejects a blank ID.

This is the XML for this operation:

<ms-bedrock:agentcore-batch-delete-memory-records
  doc:name="AgentCore batch delete memory records"
  config-ref="AWS"
  memoryId="#[payload.memoryId]">
  <ms-bedrock:delete-records>
    <ms-bedrock:memory-record-delete-input-parameters memoryRecordId="#[payload.recordId]"/>
  </ms-bedrock:delete-records>
</ms-bedrock:agentcore-batch-delete-memory-records>

Output Configuration

This operation responds with a JSON payload containing the successfulRecords and failedRecords arrays.

Configure the AgentCore - Get Memory Record Operation

The AgentCore - Get Memory Record operation retrieves a single memory record from an AgentCore Memory resource by its identifier.

To configure the AgentCore - Get Memory Record operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Memory Id

      The identifier of the AgentCore Memory resource that contains the memory record.

    • Memory Record Id

      The identifier of the memory record to retrieve.

This is the XML for this operation:

<ms-bedrock:agentcore-get-memory-record
  doc:name="AgentCore get memory record"
  config-ref="AWS"
  memoryId="#[payload.memoryId]"
  memoryRecordId="#[payload.recordId]"
/>

Output Configuration

This operation responds with a JSON payload containing the requested memory record.

Configure the AgentCore - List Memory Records Operation

The AgentCore - List Memory Records operation lists memory records in an AgentCore Memory resource based on the specified criteria. Either a namespace or a namespace path must be provided.

To configure the AgentCore - List Memory Records operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Memory Id

      The identifier of the AgentCore Memory resource for which to list memory records.

    • Namespace

      The namespace prefix to filter memory records by. Either namespace or namespace path is required.

    • Namespace Path

      The namespace path for hierarchical retrievals. Either namespace or namespace path is required.

    • Memory Strategy Id

      The memory strategy identifier to filter memory records by.

    • Max Results

      The maximum number of results to return in a single call (1-100, default 20).

    • Next Token

      The token for the next set of results, from a previous response.

    • Memory Metadata Filters

      An optional list of metadata filter expressions (1-5 expressions) to narrow the returned memory records.

This is the XML for this operation:

<ms-bedrock:agentcore-list-memory-records
  doc:name="AgentCore list memory records"
  config-ref="AWS"
  memoryId="#[payload.memoryId]"
  namespace="support/session-123"
  maxResults="20"
/>

Output Configuration

This operation responds with a JSON payload containing the list of memory record summaries and a pagination token.

Configure the AgentCore - Retrieve Memory Records Operation

The AgentCore - Retrieve Memory Records operation retrieves memory records from an AgentCore Memory resource using a semantic search over a namespace. Either a namespace or a namespace path must be provided.

To configure the AgentCore - Retrieve Memory Records operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Memory Id

      The identifier of the AgentCore Memory resource from which to retrieve memory records.

    • Search Criteria

      The semantic search criteria. The search query is required; the memory strategy ID (an opt-in filter — when omitted, records from all strategies are searched), top K (the number of top-scoring records to rank and return, 1-100, default 10), and memory metadata filters (1-5 expressions) are optional.

    • Namespace

      The namespace to search within. Either namespace or namespace path is required.

    • Namespace Path

      The namespace path for hierarchical retrievals. Either namespace or namespace path is required.

    • Max Results

      The maximum number of results to return in a single call (1-100, default 20).

    • Next Token

      The token for the next set of results, from a previous response.

This is the XML for this operation:

<ms-bedrock:agentcore-retrieve-memory-records
  doc:name="AgentCore retrieve memory records"
  config-ref="AWS"
  memoryId="#[payload.memoryId]"
  namespace="support/session-123">
  <ms-bedrock:search-criteria searchQuery="#[payload.question]" topK="5"/>
</ms-bedrock:agentcore-retrieve-memory-records>

Output Configuration

This operation responds with a JSON payload containing the matching memory record summaries and a pagination token.

Configure the AgentCore - Invoke Agent Runtime (Streaming) Operation

The AgentCore - Invoke Agent Runtime (Streaming) operation invokes an agent runtime as a Server-Sent Events (SSE) stream, piping the raw response bytes straight through to the operation payload. Response metadata (session IDs, content type, status code, and tracing headers) is surfaced as message attributes through #[attributes].

To configure the AgentCore - Invoke Agent Runtime (Streaming) operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Agent Runtime Arn

      The ARN of the agent runtime to invoke.

    • Runtime Session Id

      The identifier of the runtime session to invoke against.

    • Payload

      The JSON payload to send to the agent runtime.

    • Qualifier

      An optional runtime qualifier such as a version alias. Defaults to the DEFAULT endpoint when unset.

    • Content Type

      An optional request content type. When unset, the AWS SDK default applies.

This is the XML for this operation:

<ms-bedrock:agentcore-invoke-agent-runtime
  doc:name="AgentCore invoke agent runtime"
  config-ref="AWS"
  agentRuntimeArn="#[payload.agentRuntimeArn]"
  runtimeSessionId="#[payload.sessionId]"
  payload='#[output application/json --- { prompt: payload.question, stream: true }]'
/>

Output Configuration

This operation streams the raw SSE response bytes as the payload. The session IDs, content type, HTTP status code, and tracing headers are available through #[attributes].

Configure the AgentCore - Invoke Agent Runtime As Bytes (Non Streaming) Operation

The AgentCore - Invoke Agent Runtime As Bytes (Non Streaming) operation invokes an agent runtime and returns the raw response bytes buffered in memory. Prefer the streaming variant unless a downstream flow requires the payload to be fully buffered (for example, retry with replay). Response metadata is surfaced as message attributes through #[attributes].

To configure the AgentCore - Invoke Agent Runtime As Bytes (Non Streaming) operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Agent Runtime Arn

      The ARN of the agent runtime to invoke.

    • Runtime Session Id

      The identifier of the runtime session to invoke against.

    • Payload

      The JSON payload to send to the agent runtime.

    • Qualifier

      An optional runtime qualifier such as a version alias. Defaults to the DEFAULT endpoint when unset.

    • Content Type

      An optional request content type. When unset, the AWS SDK default applies.

    • Accept

      An optional response accept header.

This is the XML for this operation:

<ms-bedrock:agentcore-invoke-agent-runtime-as-bytes
  doc:name="AgentCore invoke agent runtime as bytes"
  config-ref="AWS"
  agentRuntimeArn="#[payload.agentRuntimeArn]"
  runtimeSessionId="#[payload.sessionId]"
  payload='#[output application/json --- { prompt: payload.question }]'
/>

Output Configuration

This operation returns the buffered response bytes as the payload. The session IDs, content type, HTTP status code, and tracing headers are available through #[attributes].

Configure the AgentCore - Stop Runtime Session Operation

The AgentCore - Stop Runtime Session operation stops an active agent runtime session. The underlying AWS API returns an empty response body; the stopped session identifier and the HTTP status code are surfaced as message attributes through #[attributes].

To configure the AgentCore - Stop Runtime Session operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Agent Runtime Arn

      The ARN of the agent runtime that owns the session.

    • Runtime Session Id

      The identifier of the runtime session to stop.

    • Qualifier

      An optional runtime qualifier such as a version alias. Defaults to the DEFAULT endpoint when unset.

    • Idempotency Token

      An opaque de-duplication identifier the AWS SDK uses to safely retry the request. The SDK auto-generates one when unset. This is not an authentication token.

This is the XML for this operation:

<ms-bedrock:agentcore-stop-runtime-session
  doc:name="AgentCore stop runtime session"
  config-ref="AWS"
  agentRuntimeArn="#[payload.agentRuntimeArn]"
  runtimeSessionId="#[payload.sessionId]"
/>

Output Configuration

This operation returns an empty payload. The stopped session identifier and the HTTP status code are available through #[attributes].

Configure the AgentCore - Get Workload Access Token Operation

The AgentCore - Get Workload Access Token operation retrieves a workload access token from Bedrock AgentCore Identity.

To configure the AgentCore - Get Workload Access Token operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Workload Name

      The workload identifier.

This is the XML for this operation:

<ms-bedrock:agentcore-get-workload-access-token
  doc:name="AgentCore get workload access token"
  config-ref="AWS"
  workloadName="#[payload.workloadName]"
/>

Output Configuration

This operation responds with a JSON payload containing the workload access token.

Configure the AgentCore - Get Resource API Key Operation

The AgentCore - Get Resource API Key operation retrieves an API key from a Bedrock AgentCore Identity credential provider.

To configure the AgentCore - Get Resource API Key operation:

  1. Select the operation on the Anypoint Code Builder or Studio canvas.

  2. In the General properties tab for the operation, enter these values:

    • Resource Credential Provider Name

      The API key credential provider name.

    • Workload Identity Token

      The workload identity token issued to the caller, typically obtained through the AgentCore - Get Workload Access Token operation.

This is the XML for this operation:

<ms-bedrock:agentcore-get-resource-api-key
  doc:name="AgentCore get resource API key"
  config-ref="AWS"
  resourceCredentialProviderName="#[payload.providerName]"
  workloadIdentityToken="#[payload.workloadIdentityToken]"
/>

Output Configuration

This operation responds with a JSON payload containing the resource API key.

See Also