Integrates with AWS Bedrock AgentCore, exposing MCP tools (list-secrets, query-db, service-execute, list-sub-tools) that agents can call while credentials remain on the customer's Akeyless Gateway. Supports AWS IAM cloud identity authentication for the AgentCore role and allows querying databases and executing services by referencing secrets without ever exposing secret values.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@akeyless-agentcore-rta query /prod/db/readonly with SELECT count(*) from orders"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Akeyless Agentic Runtime Authority (ARA / RTA) for AWS Bedrock AgentCore.
Agents call list-secrets, query-db, service-execute, and list-sub-tools. Credentials stay on your Akeyless Gateway — tool responses contain paths, metadata, and query/action results only.
Not the secrets-fetch library. For get_secret() / pulling secret values into the agent, use akeyless-agentcore-runtime. You can use both in one architecture (bootstrap via secrets runtime; DB/cloud via this package).
Concern | This package (RTA) | Secrets runtime |
Goal | Do work with secrets | Fetch secret values |
Tools | list-secrets, query-db, service-execute, list-sub-tools | list / get secret |
Credentials in agent | Never | Yes (values returned) |
Gateway | Required customer GW with ARA | SaaS API or GW OK |
Python 3.10+
Akeyless Gateway with ARA enabled (not api.akeyless.io)
Auth method your AgentCore role can use (production: AWS IAM cloud identity)
Related MCP server: openserverless-mcp
cd bedrock-agentcore-akeyless-rta
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[mcp,dev]"from akeyless_agentcore_rta import AkeylessRtaClient
client = AkeylessRtaClient() # reads AKEYLESS_* env
for s in client.list_secrets():
print(s.name, s.target_type) # never prints credentials
rows = client.query_db("SELECT 1", secret_name="/prod/db/readonly")
print(rows.results)# env (required)
export AKEYLESS_GATEWAY_URL=https://gw.example.com:8000
export AKEYLESS_ACCESS_ID=p-xxxxx
export AKEYLESS_ACCESS_TYPE=aws_iam # or access_key locally
python examples/mcp-server/server.py
# AgentCore: protocol MCP → examples/mcp-server/server.pyVariable | Required | Notes |
| Yes | Customer GW ( |
| Yes* | Auth method access id |
| No | Default |
| For access_key | Local/dev only |
(OIDC/SAML) | — | Browser login to Gateway; needs Access ID only |
| No | Default |
| No | Default |
| No | Optional default path for execute tools |
* Or AKEYLESS_TOKEN for a pre-issued session token.
Guide | Description |
Auth method, ARA RBAC, chicken-and-egg IAM bind | |
MCP protocol, envVars, invoke | |
Checklist | |
Branch protection, approvals, PyPI environment | |
Trusted publishing setup | |
Development setup |
Repository: github.com/akeyless-community/bedrock-agentcore-akeyless-rta
pip install -e ".[mcp,dev]"
pytestRun pytest inside this package directory so the workspace root does not collect unrelated tests.
Secrets at runtime (values in agent): akeyless-agentcore-runtime
Product docs: Agentic Runtime Authority
This server cannot be deployed
Governed MCP gateway: one endpoint for your tools, with credential custody and audit log.
MCP gateway with runtime security policy, tool-call-level control, and audit of agent actions.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
