Real Memory AI Agent
Real Memory AI Agent is a VS Code extension by Reallexi LLC that connects your editor to local Ollama models or the RealMemory backend. It is designed for Windows development workflows where the assistant can answer first, then plan, edit workspace files, run PowerShell commands, and summarize the result.
UX/UI & AI Director: Sam Ayoub
Publisher: ReallexiLLC
Website: https://realmemoryai.com
Repository: https://github.com/reallexi/real-memory-ai-agent
License: MIT
Screenshots
RM AI Agent chat
Session selector and menu
Key features
- Local Ollama support with
qwen3-coder:30b as the coding model.
- Planner model support with
llama3.1:8b for request understanding and implementation planning.
- RealMemory backend option with API key configuration.
- Chat-first behavior: the agent answers and explains what will happen before making code changes.
- Private planner/coder/tool logs hidden from the main chat by default.
- Activity Log menu item for detailed terminal, tool, and model activity.
- Chat sessions with a session selector, new session, rename, delete, and clear actions.
- Workspace-safe file tools: read, write, replace, list files.
- Windows PowerShell command execution from the workspace root.
- Dropdown menu with Configuration, realmemoryai.com, Activity Log, and session tools.
- VS Code status bar activity while the agent is running.
How it works
The extension uses a staged workflow:
- RM AI Agent reply — the extension first answers the user's request in the chat.
- Planner step — if the request needs code changes, the planner model turns the request into clear implementation requirements.
- Coder step — the coding model applies changes through workspace tools.
- Validation and summary — terminal/tool output is kept in the Activity Log, and the chat receives the final client-facing result.
Requirements
- VS Code
1.90.0 or newer.
- Node.js for packaging/building the extension.
- Ollama installed and running for local mode.
- Recommended local models:
ollama pull qwen3-coder:30b
ollama pull llama3.1:8b
ollama serve
Extension settings
Open Settings and search for Real Memory AI Agent, or use the extension menu: Configuration.
Important settings:
| Setting |
Description |
realMemoryAgent.backend |
Use ollama or realmemory. |
ollamaLocalAgent.baseUrl |
Ollama server URL, usually http://localhost:11434. |
ollamaLocalAgent.model |
Local coding model, usually qwen3-coder:30b. |
ollamaLocalAgent.plannerModel |
Local planner model, usually llama3.1:8b. |
realMemoryAgent.apiKey |
Bearer token for RealMemory backend. |
realMemoryAgent.showInternalActivity |
Show or hide internal logs in the chat. Default is hidden. |
realMemoryAgent.allowPowerShell |
Allow PowerShell execution in the workspace. |
realMemoryAgent.allowedTools |
Tools the agent may use. |
Usage
- Open a workspace folder in VS Code.
- Open the RM AI Agent activity bar icon.
- Ask a question or request a change.
- Use the top-right menu to open Configuration, Activity Log, website, or session actions.
Example prompt:
Install concurrently if missing. Update the root package.json to add npm run project that runs both frontend and backend. Backend must run using .\venv\Scripts\python.exe from the root venv. Modify files directly.
Commands
| Command |
Description |
RM AI Agent: Open Chat |
Opens the RM AI Agent chat view. |
RM AI Agent: Run Task |
Runs a task from an input prompt. |
RM AI Agent: Open Activity Log |
Opens the detailed output channel. |
RM AI Agent: Clear Chat History |
Clears saved chat history. |
RM AI Agent: Open Configuration |
Opens extension settings. |
RM AI Agent: Open realmemoryai.com |
Opens the product website. |
Packaging
npm install
npm run compile
npm run package
The top-level extension icon must be PNG. The Activity Bar icon may remain SVG.
Ownership
Real Memory AI Agent is published by Reallexi LLC.
UX/UI and AI direction by Sam Ayoub — UX/UI & AI Director.
