Skip to content
| Marketplace
Sign in
Visual Studio Code>Programming Languages>AgenticAgentsNew to Visual Studio Code? Get it now.
AgenticAgents

AgenticAgents

Marouan BLELLETTA

|
30 installs
| (0) | Free
Autonomous multi-agent AI coding assistant powered by Ollama — orchestrates master/worker agents to build entire projects
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

🤖 AgenticAgents (OllamaForge) v2.6.0

The Sovereign Multi-Agent Software Factory for VS Code

License: MIT VS Code Ollama Portfolio

AgenticAgents is a state-of-the-art multi-agent orchestration engine that transforms a single natural language prompt into a complete, production-ready codebase. It shifts the paradigm from simple "chat assistants" to a fully autonomous software factory that runs on your hardware or in the cloud.


🌩️ Total Architectural Freedom

  • 100% Local Sovereignty: Run entirely offline using Ollama. Your code never leaves your machine.
  • Scale to the Cloud: Seamlessly connect to Ollama Cloud (Groq, Fireworks, OpenRouter) to leverage massive 405B+ models for complex reasoning.
  • Hybrid Intelligence: Mix and match. Use a high-reasoning Cloud model as your Architect (Master) and agile local models as your Builders (Workers).

🏗️ Master/Worker Syndicate Architecture

Inspired by high-performance engineering teams, AgenticAgents utilizes a hierarchical model that outpaces traditional linear chat bots.

🧠 The Master Agent (The Architect)

The Master is the persistent brain of the mission. It maintains global project state, plans the mission, and delegates work.

  • Autonomous Discovery: Recursively explores your codebase and researches documentation via websearch.
  • Zero-Loss Succession: When context limits are hit, it writes a Testament and spawns a Successor with zero memory loss.
  • Unified Registry: All metadata, roadmaps, and logs are consolidated in a clean .agenticagents directory.

👷 The Worker Agents (The Builders)

Workers are specialized, short-lived agents spawned to own specific files or technical tasks.

  • Massive Parallelism: Spawns concurrent workers with rolling replenishment (up to 500 tasks).
  • Hardened Specialized Tools: Dedicated handlers for Python, Java, Maven, npm, and Shell with strict whitelisting.
  • Self-Healing Loop: Uses the problems tool to detect VS Code build errors and fix them autonomously.

🎯 Three Specialized Mission Modes

Mode Strategy Phase Flow
ASK Analytical Discovery Discovery → Iterative Research → Direct Answer
FAST Direct Implementation Workspace Analysis → Immediate Implementation
PLAN-BUILD Phased Orchestration Discovery → Technical Roadmap (todo.md) → Parallel Building

🚀 Professional-Grade Capabilities

👁️ Intelligent Vision

Drop a UI mockup or whiteboard sketch. The Vision Integration automatically activates during planning to analyze designs and translate them into code.

🌐 Real-Time Web Research

Agents don't just guess. Integrated websearch and webfetch allow them to browse the latest documentation and research solutions in real-time.

⏲️ True Temporal Rollback

Did the AI take a wrong turn? Hover over any query and click Cancel or Modify. The forge instantly reverts all filesystem creations, edits, and deletions back to the precise moment before that query was issued.

🔄 Dynamic Interjection & Pivot

Don't wait for completion. Send Interjective Guidance mid-workflow. The Master Agent will immediately abort its current thought process to pivot based on your feedback.

📊 Change Management UI

  • Pending Changes Panel: Review every single file change in a dedicated UI sidebar.
  • Visual Diff Sync: Approve or revert changes with full diff visibility. Approvals automatically clean up the temporary editor tabs.

🛠️ How to Launch a Mission

  1. Open the Forge: Press Ctrl+Shift+A or click the sidebar icon.
  2. Define the Goal:

    "Refactor this React app to use clean architecture and add a gRPC backend in Go."

  3. Review the Discovery: Watch the Master explore your files and design screenshots.
  4. Approve the Roadmap: The agents start building according to the .agenticagents/todo.md.
  5. Monitor the Factory: Watch the Agent Monitor track tool calls and file edits in real-time.
  6. Finalize: Review changes in the Pending Changes UI and approve the final build.

👨‍💻 Connect with the Architect

Built with ❤️ by Marouan BLELLETTA.
Interested in collaborating, reporting bugs, or hiring?

Portfolio

100% Privacy | 100% Autonomous | 100% Engineering
  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft