كيفية استخدام أكثر من وكيل ذكاء اصناعي على نفس المهمة في نفس الوقت

Ruflo — How to Turn VS Code into a Full AI Agent Orchestration System

I recently came across a promising project called Ruflo with around 55,000 stars on GitHub, and I found it potentially very useful — especially for those who rely on building projects that depend fully or partially on Claude or other AI providers.

What is Ruflo?

If you use Claude Code daily in VS Code, you’re familiar with its core limitations: every session is isolated, no memory, no coordination, and every task starts from scratch.

Ruflo (formerly known as Claude Flow) solves this problem significantly. More than that, it’s an AI agent orchestration platform built on top of Claude Code, adding:

• 98 specialized agents (security, testing, documentation, architecture…)
• Memory that learns and remembers across sessions
• A swarm of up to 15 agents working in parallel on the same task simultaneously, each in its own role.
• Continuous background operation — can run on your local or private server.

The project is open source, has received over 50,000 stars on GitHub, and is still under active development.
This is in addition to the ability to integrate and connect most language models. You can also customize agents based on your needs and budget. For example, you can wake up large, expensive models for complex tasks, while leaving smaller, cheaper models to handle simpler ones.
You can read more about the project here: https://github.com/ruvnet/ruflo

Requirements — Make Sure You Have:

• Node.js installed (verify with node --version)
• VS Code
• A claude.ai subscription (Pro or higher)

Step 1 — Install Claude Code CLI

Open the Terminal in VS Code and run:

npm install -g @anthropic-ai/claude-code

After installation completes, verify it:

claude --version

You should see something like:

2.1.141 (Claude Code)

Step 2 — Launch Claude Code and Sign In

claude

A login screen will appear — choose whichever method suits you. I personally use option 1 with my Claude subscription:

Select login method:
1. Claude account with subscription · Pro, Max, Team, or Enterprise
2. Anthropic Console account · API usage billing
3. 3rd-party platform

The browser will open automatically — sign in with your claude.ai account, then copy the short code that appears and paste it in the Terminal after the > prompt.

Step 3 — Claude Code is Ready!

After successfully signing in, the main screen appears:

Claude Code v2.1.141
Sonnet 4.6 · Claude Pro
✓ Anthropic marketplace installed

Now exit temporarily:

/exit

Step 4 — Install Ruflo in Your Project

Navigate to your project folder and run:

cd E:\YOUR-PROJECT\FOLDER
npx ruflo@latest init

When asked Ok to proceed? (y), type y and press Enter.

After installation, you’ll see:

RuFlo V3 initialized successfully!
+--------- Summary ----------+
| Directories: 11 created    |
| Files: 85 created          |
+----------------------------+
| Agents:  98 agents         |
| Commands: 10 commands      |
+----------------------------+

Step 5 — Start the Daemon (Background Workers)

npx ruflo daemon start

Type y to confirm. Result:

[OK] Daemon started in background (PID: 20172)
[INFO] Logs: .claude-flow\daemon.log

This means Ruflo is now running in the background, continuously monitoring your project.

Step 6 — Initialize Memory

npx ruflo memory init

Expected result:

+-------------------- Configuration --------------------+
| Backend:           hybrid                             |
| Vector Embeddings: ✓ Enabled                          |
| Pattern Learning:  ✓ Enabled                          |
| HNSW Indexing:     ✓ Enabled                          |
+-------------------------------------------------------+
Verification passed (6/6 tests)

This memory will learn from every task you run and recall it in future sessions.

Step 7 — Initialize the Swarm (All Agents Together)

npx ruflo swarm init

You’ll be asked to choose a swarm topology — select Hierarchical Mesh (the last option) using the down arrow, then press Enter.

Selected: Hierarchical Mesh
+------------+----------------------------+
| Swarm ID   | swarm-xxxxxxxxxxxxxxx      |
| Topology   | hierarchical-mesh          |
| Max Agents | 15                         |
| Auto Scale | Enabled                    |
+------------+----------------------------+
[OK] Swarm initialized successfully

A question might come to mind: why only 15 agents when there are 98?

The short answer:

The 98 agents are the full library of specializations stored as files in .claude/agents/ — they’re dormant and consume no resources until needed.

The 15 agents are the maximum that can run in parallel at the exact same moment on the same task!

When you give a command, Ruflo selects only what it needs from the 98 and runs them together. For example, if you ask “review the project”:

Security Agent + Code Agent + Tests Agent
+ Documentation Agent + Performance Agent
= 5 agents working in parallel at the same moment

How to Use It?

I personally use it through the direct chat interface in VS Code — I just type my request and Ruflo responds based on the shared project context. You can also use the Terminal directly if you prefer!

Note: This project is currently in an alpha stage (v3.7.0-alpha). It’s stable enough for daily development use, but I wouldn’t recommend relying on it fully in sensitive production environments until it reaches a stable release.

Conclusion

This has been my modest experience with running multiple agents simultaneously on the same request. I have prior experience with OpenClow on a local VPS server and running multiple models locally at no cost — that might be worth exploring too.

Personally, I rely entirely on Claude and Replit for my projects, but for projects that require extensive review and security testing, I use multiple agents for experimentation, then take my findings to the environment where I actually build the project.