AWS adds a guide for migrating knowledge bases to Amazon Bedrock Managed Knowledge Base to the Agent Toolkit for AWS
AWS has added a migration guide to the Agent Toolkit for AWS for moving a knowledge base from customer-managed to Amazon Bedrock Managed Knowledge Base. A coding agent proposes the migration plan; the managed version offers higher quotas, automatic management, and a different, usage-based pricing model.
Amazon Web Services has added a migration guide to the Agent Toolkit for AWS that helps with moving from a customer-managed knowledge base to an Amazon Bedrock Managed Knowledge Base. The guide is part of the amazon-bedrock skill and gives the coding agent the context needed to plan the migration together with the user. The agent first performs only a read-only survey of the existing configuration and proposes a plan, which the user approves step by step. After approval, the agent either creates a new managed knowledge base and its data sources alongside the original one, or generates a migrated infrastructure-as-code template for the company's pipeline. The user still has to carry out the data ingestion and adjust the Retrieve call in the application themselves; according to the company, the agent can only guide them through this.
According to AWS, Managed Knowledge Base takes over ingestion, storage, indexing, and retrieval, and automatically scales based on the volume of content. Hybrid search (a combination of keyword and semantic search) is always enabled without the need for index configuration; the service includes management of the embedding model, a reranker, and a built-in parser for multimodal file types. A custom embedding model can still be used when creating a knowledge base. Features such as agentic retrieval, integration with AgentCore Gateway, and resource-based policies are available only in the managed version and are missing from the customer-managed knowledge base.
The new default quotas for the managed version are significantly higher: 10 000 knowledge bases per account and region (versus 100 for customer-managed), 200 data sources per knowledge base (versus 5), 50 concurrent ingestion jobs (versus 1), and 10 000 characters per input query (versus 1 000).
According to the source, pricing models differ: for a customer-managed knowledge base, you pay separately for the chosen vector store, and some stores have a minimum operating cost even at low usage. In the managed version, storage is included in the base rate of 5 USD per GB per month, standard search costs 1 USD per 1 000 requests including reranking, embedding with the default model is included, and agentic retrieval costs an additional 4 USD per 1 000 queries on top of the regular search costs. For comparison, according to AWS, building a custom agentic workflow on a customer-managed knowledge base costs approximately 18 to 24 USD per 1 000 queries before accounting for custom development and maintenance, with the price varying depending on model token usage. The source article contains further details on cost items that are not fully available here.
Why it matters
Companies running knowledge bases for RAG applications on Amazon Bedrock can use the tool to assess whether switching to the managed variant is worthwhile for them – the decision depends on the balance between the level of control over the vector store, the cost at low and high usage, and the availability of features such as agentic retrieval. For developers, the guide saves time when planning the migration, since the agent performs a read-only analysis and proposes steps for approval, but they still have to complete the actual data ingestion and application code changes themselves.
Two audiences, two different impacts
What this means
For individuals
A developer managing a Bedrock knowledge base no longer has to build the migration plan manually – a coding agent performs a read-only analysis, proposes a plan for approval step by step, and either creates a new managed knowledge base or generates an IaC template; the developer still has to handle the data ingestion and adjust the Retrieve call in the application themselves.
For a business
Companies running knowledge bases on Amazon Bedrock now have a new tool for planning migration and a fundamentally different pricing model (pay for actual usage instead of operated infrastructure), higher quotas (10 000 knowledge bases per account/region versus 100, 200 data sources versus 5), and access to features such as agentic retrieval or AgentCore Gateway, which…
DevelopmentCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.