The Amazon Bedrock console: a 2026 walkthrough.

IWhat the Bedrock console is, and what it is not

The console is the AWS Management Console surface for Amazon Bedrock. Every click maps to a Bedrock API action, which makes it a good way to learn the service.

Amazon Bedrock is a managed, multi-model inference service: one API for Claude, Llama, Mistral, Amazon Nova, Cohere, Stability AI, AI21, DeepSeek and others, with no GPUs to run (overview at Amazon Bedrock). The console is the control panel: try models, configure the managed building blocks (RAG, agents, safety, orchestration), and inspect costs.

The console is good at setup, exploration and configuration. Comparing two models on a prompt, pointing a Knowledge Base at a folder, drafting a Guardrail or running a small evaluation is faster by clicking, and each action is an auditable AWS call.

It is not your production runtime. The console produces resources (a Knowledge Base ID, an Agent alias, a Guardrail ID, a Flow) that your code then calls through the API. Section VIII draws that line.

IIThe navigation map: every panel and what it does

AWS renames and regroups panels over time, so treat this as a map of jobs, not a screenshot.

IIIModel access and the model catalog

Older guides say Bedrock starts with every model switched off. That changed. In commercial Regions AWS now enables access to many models by default.

IVThe Playgrounds: chat, text and image/video

The Playgrounds let you prompt a model in the browser, adjust parameters and compare models. Every call is billed per token.

The Chat / Text Playground takes a prompt (and optional system prompt) and lets you tune temperature, top-p, max tokens and stop sequences. Its most useful feature is compare mode: send one prompt to two models and read the answers, latency and token use side by side.

The Image / Video Playground does the same for media models (Amazon Nova Canvas, Stability, Amazon Nova Reel): prompt, size, variants, seed and style.

Most Playgrounds can show the API request behind your session. Prototype the prompt and parameters by hand, then copy the request into your code.

VThe builders: Knowledge Bases, Agents, Guardrails and Flows

Each builder has a guided setup in the console. You configure a real resource, test it in place, and get an ID your app calls.

The builders compose: a typical assistant is a Knowledge Base for grounding, wrapped in a Guardrail, exposed through an Agent, and maybe orchestrated by a Flow. All of it can be configured and tested before you write a serving loop.

• Knowledge Bases: Point at a data source (often S3), pick an embeddings model and a vector store, then sync. Bedrock chunks, embeds and indexes the documents. Test it in the console and see answers with citations. Your app calls RetrieveAndGenerate or Retrieve. See Bedrock Knowledge Bases.
• Agents: Write instructions, pick a model, and define action groups backed by Lambda or an OpenAPI schema. Optionally attach a Knowledge Base and a Guardrail. Test in the console and watch the trace. Your app calls InvokeAgent on an alias. See Amazon Bedrock Agents.
• Guardrails: Set content filters, denied topics, word filters, PII handling and grounding checks. Version the policy and apply its ID to model calls, Agents or Knowledge Bases. See Bedrock Guardrails.
• Flows: A canvas where you wire prompts, models, Knowledge Bases, Agents and logic into a versioned workflow. Your app calls InvokeFlow on an alias. See Bedrock Flows.

VIEvaluations, Prompt Management and custom models

Three panels turn a prototype into something you can measure, reuse and specialize.

VIIBilling, IAM and governance

Who can do what (IAM) and what it costs (Billing) govern everything you click.

Cost: Bedrock charges flow into standard AWS billing. Use Cost Explorer, filter to Amazon Bedrock, and group by usage type or model. Console activity is billable: Playground prompts, evaluation jobs and Knowledge Base syncs and queries all consume tokens or resources. A long afternoon comparing frontier models adds up.

Access: each console action is a Bedrock API call governed by IAM. A teammate who sees errors or empty panels usually lacks permissions such as bedrock:ListFoundationModels, the invoke actions, or the builder actions. First use of a third-party model also needs Marketplace permissions.

Two settings to turn on early: model invocation logging (request and response payloads to S3 or CloudWatch) and, at organization level, SCPs that deny invoking models you have not approved.

VIIIConsole vs API: what you can do click-only, and when to switch

If a person does it once, use the console. If software does it repeatedly, use the API.

Trying and comparing models, building and testing a Knowledge Base, Agent, Guardrail or Flow, running an evaluation, drafting prompts and starting a fine-tune all fit the console. Serving user traffic, batch jobs, CI/CD and infrastructure as code belong in the API and IaC.

For the API itself see the Bedrock API reference. For cost levers at volume see Bedrock pricing, prompt caching and batch inference.

IXWho can fund the move from console prototype to production

Console experiments and production traffic both draw down AWS credits automatically.

Partner-filed funding: a vetted AWS partner can file POC (proof of concept) funding to take a prototype to a working pilot, MAP funding if you are moving from another cloud, or incremental workload (IW) funding if you already run on AWS and this is new work. For startups this is typically $50K to $150K. Depends on the workload; AWS approves.

AWS Activate: you apply yourself. Activate Founders gives self-funded startups $1,000 to start, up to $5,000. Activate Portfolio gives up to $200,000 with the Org ID from your VC, accelerator or angel group, if it is an Activate Provider. A consulting partner cannot file Activate for you.

CloudRoute matches you with one vetted AWS partner who can turn the console prototype into a governed production app and check which funding fits. Startups pay CloudRoute nothing. Partners pay CloudRoute a monthly membership plus 10% of AWS-approved funding.