ISV Agent ForgeBerlin (BER21) · Thu 16 July 2026
Copy-paste ready

Tooling cheat sheet

Get productive fast — environment, your agentic IDE on your laptop, and a Strands quickstart.

Before you arrive (on your laptop)

The afternoon fork runs on your own laptop, so install ahead of time:

curl -LsSf https://astral.sh/uv/install.sh | sh # uv (Python package + venv manager)

First 10 minutes (workshop environment — for the morning labs)

Labs 1–2 run in VS Code Server on EC2, region us-west-2.

  1. Open the Workshop Studio event dashboardEvent Outputs → open the Code Editor URL (VS Code Server).
  2. In the VS Code terminal, install dependencies and set the region:
uv sync aws configure set region us-west-2 aws configure get region
  1. Open 01-bedrock-knowledge-base/01_create_knowledgebase.ipynbSelect KernelPython EnvironmentsVenvUse Existing (.venv).
#1 snag: if the notebook can't find the kernel, run source .venv/bin/activate in the terminal, then re-select the kernel.

Your agentic IDE (afternoon fork — on your laptop)

Build your agent in your own IDE against your workshop AWS credentials. Get those from the Workshop Studio dashboard, then point your laptop at us-west-2:

# Workshop Studio dashboard -> "Get AWS CLI credentials" -> paste the export block, then: aws configure set region us-west-2 aws sts get-caller-identity # confirms you're in the workshop account

Kiro

kiro-cli chat
  • Steering: .kiro/steering/*.md (or AGENTS.md)
  • Skills: .kiro/skills/NAME/SKILL.md
  • Powers: MCP + knowledge bundles (Kiro panel)
  • MCP: .kiro/settings/mcp.json · Agents: .kiro/agents/
  • Ask Kiro about Kiro: /guide

Claude Code

npm install -g @anthropic-ai/claude-code
  • Memory: CLAUDE.md (read every session)
  • Skills: .claude/skills/NAME/SKILL.md
  • Subagents: .claude/agents/
  • MCP: claude mcp add ... · Hooks: .claude/settings.json
  • Slash commands: .claude/commands/*.md
Write once, use in both. Skills follow the open Agent Skills standard — a SKILL.md works in Kiro and Claude Code. And CLAUDE.mdAGENTS.md ≈ Kiro steering.

Recommended setup for this build

Install Strands (Python)

uv add strands-agents strands-agents-tools

Add an AWS docs MCP server

Kiro.kiro/settings/mcp.json:

{ "mcpServers": { "aws-docs": { "command": "uvx", "args": ["awslabs.aws-documentation-mcp-server@latest"] } } }

Claude Code — one command:

claude mcp add aws-docs -- uvx awslabs.aws-documentation-mcp-server@latest

Starter project context (steering / CLAUDE.md / AGENTS.md)

# Building a Strands (Python) agent on Amazon Bedrock AgentCore. - Region: us-west-2. Package manager: uv. Python 3.10+. - Framework: Strands Agents (model-driven). Model: Claude Sonnet 5 (us.anthropic.claude-sonnet-5). - Pattern: Agents-as-Tools — a supervisor routes to specialist agents. - Deploy target: AgentCore Runtime. External APIs via AgentCore Gateway (MCP). - Always: write tests, least-privilege IAM, no secrets in code.

Strands quickstart (a tool-using agent in ~10 lines)

from strands import Agent, tool @tool def get_current_rate(term_years: int) -> str: "Return today's rate for a given mortgage term." return lookup_rate(term_years) agent = Agent( model="us.anthropic.claude-sonnet-5", # current Sonnet for your own build system_prompt="You are a helpful mortgage assistant.", tools=[get_current_rate], ) print(agent("What are 15-year rates today?"))
Model note: the morning labs are pinned to Claude 3.7 Sonnet for reproducibility. For your own agent use a current model — Claude Sonnet 5 (us.anthropic.claude-sonnet-5) or the prior gen us.anthropic.claude-sonnet-4-6. Swapping is a one-line model= change.
Strands docs →
Quickstart, tools, multi-agent, model providers.
AgentCore docs →
Runtime, Gateway, Identity, Memory, Observability, Tools.
MCP →
The open tool protocol Gateway speaks.
Agent Skills →
Portable SKILL.md across Kiro & Claude Code.
← Agenda.