Amazon DynamoDB

Read, write, query, and scan Amazon DynamoDB tables from automated flows. Agents persist workflow data, run PartiQL statements, and batch-load items with values returned as plain JSON.

10 actions available

Upstream SQS receive returns a batch of queued events

Agent maps each event into a DynamoDB item shape

Agent calls Describe Table to confirm the target table is Active and the schema matches

Agent calls Batch Write Items with up to 25 items per request, auto-retrying unprocessed items

Agent confirms the success-count against the input count

Operations channel gets the batch summary

Persistent unprocessed-item runs or schema mismatches are routed to the on-call engineer for review

What This Integration Enables

Agents perform single-item reads and writes (Get Item, Put Item, Update Item, Delete Item) and run keyed Queries or table Scans with filters and pagination. Batch actions move up to 100 items per Batch Get and up to 25 per Batch Write, automatically retrying anything the service marks as unprocessed. PartiQL via Execute Statement lets a workflow run SQL-shaped statements when that is the right tool. Describe Table exposes a table's primary key, indexes, and current state so the agent can confirm shape and readiness before writing. The connector covers the read/write surface a workflow needs from DynamoDB; the surrounding flow decides what gets persisted, queried, or escalated.

Without FlowRunner

Custom marshalling code Every DynamoDB write requires hand-maintained type-mapping code on the caller

Pagination glue per query Each query implementation reinvents cursor handling and unprocessed-item retry

Inconsistent cross-account access One-off scripts hold IAM credentials and rotate inconsistently

With FlowRunner

Plain-JSON I/O Items are supplied and returned as plain JSON; the connector handles typed attribute-value conversion

Cursors and retries handled Query, Scan, and batch actions return opaque cursors and auto-retry unprocessed items

One configuration for both auth modes API Key and IAM Role are selected per-flow; the same connector serves single- and cross-account writes

Use Case Scenarios

Queue-Backed Event Persistence

The agent calls Receive Messages on an SQS queue, maps each event into a DynamoDB item shape, and calls Batch Write Items to persist up to 25 events per request. The connector handles unprocessed-item retries automatically. The agent then calls Delete Message with the receipt handles so the events are not redelivered. The queue stops growing, the table stays current, and the engineer who used to babysit the writer can work on something else.

Key-Conditioned Lookup and Fan-Out

A flow needs to notify subscribers whenever a record matching a key condition lands. The agent calls Query against the primary index with the condition, paginates through results using the returned cursor, and for each item calls Publish Message on an SNS topic so downstream subscribers receive the notification. The same pattern works against a Global Secondary Index by passing the GSI name in the Query parameters.

Schema-Aware Reporting Export

A monthly reporting flow scans a table, paginates the results, and writes them to a reporting spreadsheet. The agent first calls Describe Table to confirm the schema, then calls Scan with a filter expression. Each page of results goes to Google Sheets via Add Rows. When the cursor returns null, the export is complete and the agent posts a summary in the operations channel.

Human-in-Loop Highlight

A persistent failure in Batch Write Items, where the same items keep coming back unprocessed across multiple retries, is a signal that something structural is wrong, not a transient throttle. The agent pauses, posts the unprocessed-item count and the table name to the on-call channel: "Batch Write Items returned the same [n] unprocessed items across [retries] retries against table [name]. Likely capacity exhaustion or schema mismatch. Continue retrying, drain to a dead-letter queue, or pause the flow?" The engineer decides. The agent does the durable thing: it stops writing rather than burning capacity in a loop.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

10 actions

Single-Item I/O

Get Item Retrieves a single item by primary key as plain JSON. Used to look up a record before deciding what to do with it.
Put Item Inserts or replaces an item in a table from plain JSON. Used to persist a workflow output or an event payload.
Update Item Updates specific attributes on an item. Used either with a simple Updates object or a raw Update Expression for ADD, REMOVE, or conditional math operations.
Delete Item Deletes a single item by primary key. Used in cleanup and tombstone-marking flows.

Read at Scale

Query Queries a table or index with key conditions and filters. Used when the access pattern matches the partition-key shape. Returns an opaque cursor for the next page.
Scan Scans a table for items matching a filter. Used in export, audit, and migration workflows where the access pattern does not fit the key shape. Returns an opaque cursor for the next page.
Execute Statement (PartiQL) Runs a SQL-compatible PartiQL statement against DynamoDB. Used when a SQL-shaped query is the right tool for the read.

Batch I/O

Batch Get Items Retrieves up to 100 items at once by primary key. Auto-retries any keys the service marks unprocessed.
Batch Write Items Puts or deletes up to 25 items in a single request. Auto-retries any items the service marks unprocessed.

Inspection

Describe Table Inspects a table's primary key, indexes, and current state. Used to validate shape and readiness before reading or writing.

Start building with Amazon DynamoDB

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us