AWS Bedrock

Run inference against Amazon Bedrock foundation models from Anthropic, Amazon, Meta, Mistral, Cohere, and Stability AI through one connector, and discover which models your account can call.

5 actions available

New support case created in the queue

Agent sends the case text to Invoke Model for classification

Model returns the category, urgency, and a one-line summary

Agent routes low-urgency cases automatically by category

Agent assigns the case to the right team

Team channel gets the summary and category

High-urgency or ambiguous cases route to a lead before assignment

What This Integration Enables

Bedrock gives an agent access to a catalog of foundation models without leaving AWS. The connector invokes a model, streams or returns the completion, and lists the models your account can call, so inference stays inside the account, region, and IAM boundaries your security team already signed off on. For an AWS-native operation, that removes a common objection to putting a model in a production flow: the data never leaves the tenant. What makes it operational rather than experimental is the layer around the call. An orchestration layer routes the model's output into the next system and pulls a person in when the stakes warrant it, and FlowRunner is built for that layer.

Without FlowRunner

Models outside AWS Inference runs on a separate vendor, outside existing governance

One model, hard-coded Switching foundation models means a new integration

Cases sorted by hand Support staff read and route every incoming case

With FlowRunner

Inference inside AWS Model calls run under the same account and region as your data

Any Bedrock model Anthropic, Amazon, Meta, and more behind one connector

Cases pre-sorted The model categorizes and routes before anyone opens the queue

Use Case Scenarios

Support Case Classification

Every new support case needs a category, an urgency read, and a team. The agent sends the case text to a Bedrock model and gets back a structured classification. Low-urgency cases route automatically by category. The model handles the sorting that used to consume the first hour of every support shift, and the queue is triaged before anyone opens it.

Internal Knowledge Answers

Employees ask the same policy and process questions repeatedly. The agent retrieves the relevant internal documents, passes them to a Bedrock model with the question, and returns a grounded answer inside AWS so no internal content leaves the account. Answers that the model is confident about return directly; anything touching a sensitive policy is routed to the owning team instead of answered automatically.

Document Summarization at Scale

Long reports pile up faster than anyone reads them. The agent sends each report to a Bedrock model for a structured summary with key figures and risks, then files the summary alongside the original. Reviewers read a short brief with the source one click away. When the model flags a figure it is unsure about, the summary marks it for a person rather than presenting it as fact.

Human-in-Loop Highlight

A model running inside your AWS account is still a model, and its output is still an input to a decision, not the decision itself. When a Bedrock call produces a classification, an answer, or a summary that drives a real action, FlowRunner routes it through a [human-in-the-loop](/concepts/human-in-the-loop/) step: the agent pauses, assembles the model's output and the context behind it, and sends it to the right person via Slack or email. They decide. The model handles the volume; a person owns the judgment calls.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

5 actions

Inference

Converse Sends messages to a Bedrock foundation model through the unified Converse API, which works identically across Anthropic Claude, Amazon Nova/Titan, Meta Llama, Mistral, Cohere, and other chat models.
Invoke Model Invokes a Bedrock model directly with a raw, model-specific request body and returns the parsed model-specific response.
Generate Image Generates an image from a text prompt using an image model such as amazon. titan-image-generator-v2:0 or a Stability AI model, then saves the decoded PNG to FlowRunner file storage and returns a downloadable URL.

Models

List Foundation Models Lists the Amazon Bedrock foundation models available in the configured region, with optional filtering by provider name or output modality.
Get Foundation Model Retrieves the full details of a single Amazon Bedrock foundation model by its model ID, including provider, supported modalities, streaming support, inference types, and lifecycle status.

Start building with AWS Bedrock

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

Start Building Free Talk to us