About cloudkb
cloudkb is an MCP server for Amazon Bedrock users to inspect knowledge bases and data sources, manage ingestion jobs, and retrieve information. It covers control-plane and runtime data-plane requests, but excludes knowledge base lifecycle operations such as create, update, and delete.
Key features
- 01List and inspect knowledge bases and data sources
- 02View ingestion jobs
- 03Start or stop ingestion jobs
- 04Retrieve and retrieve with generation
Use cases
- 01Inspect Bedrock knowledge bases and data sources
- 02Review or manage ingestion jobs
- 03Run retrieval or retrieval with generation against a knowledge base
MCP support
It exposes 10 tools as a stdio MCP server for Bedrock Agent control-plane and Agent Runtime data-plane requests. Calls require AWS credentials; SigV4 signing must be supplied through an injected signer or executor.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
10Saved tool list · updated 1w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| next_token | string | No | Pagination token returned by a previous list call to fetch the next page. |
| max_results | number | No | Maximum number of knowledge bases to return in one page. |
Details
From the project READMEcloudkb
Amazon Bedrock knowledge bases as MCP tools. cloudkb is a clean-room rebuild of the core surface of the AWS Bedrock knowledge base MCP server, under a new name: the same tool names and documented REST request shapes, fresh code, no copied branding.
Tools (core subset, 10)
list_knowledge_bases, get_knowledge_base, list_data_sources,
get_data_source, list_ingestion_jobs, get_ingestion_job,
start_ingestion_job, stop_ingestion_job, retrieve,
retrieve_and_generate.
The first eight tools shape Bedrock Agent (control plane) requests;
retrieve and retrieve_and_generate shape Bedrock Agent Runtime
(data plane) requests. Request shaping is real and unit-tested; every
call goes through an injectable executor interface.
Not in this subset: knowledge base and data source create, update and delete lifecycle, chunking configuration, agents and aliases, flows, prompts, guardrails, evaluation jobs, model invocation and sessions.
Setup
export AWS_ACCESS_KEY_ID=...
export AWS_SECRET_ACCESS_KEY=...
export AWS_REGION=us-east-1 # optional, this is the default
Without credentials, every tool call fails with an error naming
AWS_ACCESS_KEY_ID.
Signing gap, stated plainly: SigV4 signing is not implemented in
this rebuild. Signing is delegated to an injectable signer/executor
hook; with credentials present but no signer injected, a call fails
with SigV4 signing not configured: inject a signer/executor instead
of sending an unsigned request.
CLI
npx @reneturcios/cloudkb list_knowledge_bases '{}'
npx @reneturcios/cloudkb retrieve '{"knowledge_base_id":"KB1","query":"refund policy"}'
npx @reneturcios/cloudkb mcp # stdio MCP server
Each tool name is a subcommand taking one JSON object argument.
MCP client config
{
"mcpServers": {
"cloudkb": {
"command": "npx",
"args": ["-y", "@reneturcios/cloudkb", "mcp"],
"env": { "AWS_ACCESS_KEY_ID": "...", "AWS_SECRET_ACCESS_KEY": "...", "AWS_REGION": "us-east-1" }
}
}
}
Docker
docker run -i -e AWS_ACCESS_KEY_ID -e AWS_SECRET_ACCESS_KEY ghcr.io/cloudcorprecords/cloudkb
Development
npm install
npm run build
npm test
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.