agentic-flow

🥈Silver

Easily switch between alternative low-cost AI models in Claude Code/Agent SDK. For those comfortable using Claude agents and commands, it lets you take what you've created and deploy fully hosted agents for real business purposes. Use Claude Code to get the agent working, then deploy it in your favorite cloud.

592960Updated 1mo ago

Intermediate30min to implementautomation

Saves ~50 min per use

Quick InstallView Source

git clone https://github.com/ruvnet/agentic-flow.git

Works with:

Claude

Overview

About This Skill

https://www.npmjs.com/package/agentic-flow

How to Use

[{"step":"Identify your current model and target model. Run `claude agents list` to see available models in your workspace. Note the model names (e.g., `claude-3-5-sonnet-20241022` vs. `claude-3-haiku-20240307`).","tip":"Use `claude agents switch --help` to explore additional flags like `--temperature` or `--max-tokens` for fine-tuning."},{"step":"Prepare your agent’s task description. Write a concise 1-2 sentence summary of what the agent does (e.g., \"Process customer support tickets and route them to the appropriate team based on sentiment analysis.\"). Save this in a file named `task_description.txt` for reuse.","tip":"Include edge cases in your task description (e.g., \"Handle tickets with no clear sentiment by escalating to a human.\")."},{"step":"Execute the switch command. Use the template: `claude agents switch --from [CURRENT_MODEL] --to [TARGET_MODEL] --task \"[YOUR_TASK_DESCRIPTION]\"`. For example: `claude agents switch --from claude-3-5-sonnet-20241022 --to claude-3-haiku-20240307 --task \"Route support tickets by sentiment and urgency\"`.","tip":"Add `--dry-run` to preview changes without applying them first."},{"step":"Validate the agent’s performance. Run a test on your dataset using `claude agents test --input [TEST_DATA_FILE] --output [RESULTS_FILE]`. Compare the results against your baseline (e.g., previous model’s accuracy or cost).","tip":"Use `claude agents metrics` to log performance data over time and track improvements."},{"step":"Deploy the updated agent. If using Claude Agent SDK, run `claude agents deploy --cloud [CLOUD_PROVIDER] --region [REGION]` (e.g., `--cloud aws --region us-east-1`). For Claude Code, export the agent as a Docker container and push to your registry.","tip":"Set up a CI/CD pipeline (e.g., GitHub Actions) to automate model switching and testing for future updates."}]

Use Cases

Deploy self-learning AI agents to automate code review processes and improve code quality.

Utilize multi-agent coordination to optimize project planning and resource allocation in software development.

Implement real-time adaptive learning agents for dynamic customer support solutions.

Create intelligent testing agents that learn from past failures to generate comprehensive test cases.

Setup & Installation

Quick Install

No install command available. Check the GitHub repository for manual installation instructions.

Alternative Install (Git Clone)

git clone https://github.com/ruvnet/agentic-flow

Requirements

Claude Code or compatible AI agent
Works with: Claude

Quick Start Guide

Install the Skill

Copy the install command above and run it in your terminal.

Open Your AI Agent

Launch Claude Code, Cursor, or your preferred AI coding agent.

Try It Out

Use the prompt template or examples below to test the skill.

Customize

Adapt the skill to your specific use case and workflow.

Usage Examples

Prompt Template

Switch from [CURRENT_MODEL] to [TARGET_MODEL] in my Claude Code/Agent workflow for [TASK_DESCRIPTION]. Ensure the agent's logic remains intact while optimizing for cost and performance. Use the following command structure: `cd /path/to/project && claude agents switch --from [CURRENT_MODEL] --to [TARGET_MODEL] --task "[TASK_DESCRIPTION]"`. After switching, validate the agent's output with a test run on [TEST_DATA_FILE].

Example Output

I’ve successfully switched your agent from **Claude 3.5 Sonnet** to **Claude 3 Haiku** for the **customer support ticket triage workflow** while preserving the original logic. The agent now processes tickets with a 40% reduction in cost per request while maintaining 95% accuracy on the test dataset (`tickets_test_set.json`). Here’s the validation breakdown:

- **Processing Speed**: Reduced from 1.2s to 0.8s per ticket (33% faster).
- **Cost Savings**: $0.002 per ticket (vs. $0.0035 with Sonnet).
- **Accuracy**: 95% match against human-labeled tickets (vs. 96% with Sonnet).
- **Edge Cases**: Handled 12% more ambiguous queries (e.g., "The service is slow" vs. "I’m angry") by falling back to a rule-based fallback.

The agent is now deployed in your AWS Lambda function (`support-triage-haiku`) and is ready for production traffic. Monitor the CloudWatch logs for the next 2 hours to confirm stability.