Officeless Code Intelligence — The Officeless platform for AI-assisted coding

The Officeless platform for AI-assisted coding

Officeless Code Intelligence connects Mekari developers to safe, productive AI inside VS Code — with a local macOS runtime, guardrails aligned to Officeless standards, MCP tools for platform APIs, and releases you can trust.

Scroll for a guided tour: macOS bridge → VS Code extension → Code Intelligence IDE → guardrails → downloads.

Explore the platform Download macOS app

Local-firstOfficeless guardrailsOfficeless MCPVS Code & branded IDEDevOps Central profile

Platform

One stack for Officeless AI coding

Officeless Code Intelligence is Mekari's internal platform for building with AI — not a single app, but a connected set of runtimes, editor surfaces, guardrails, and release channels. Start with the overview below, then read each layer in order.

01
macOS bridge
A lightweight menu bar agent exposes a localhost API. Apple Intelligence, Ollama, or cloud providers route through one control plane on your Mac.
Read more →
02
VS Code extension
Copilot-style sidebar chat, inline completion, and auto-apply — workspace-aware, with general or DevOps Central profiles.
Read more →
03
Code Intelligence IDE
A branded VS Code branded IDE build for Officeless teams — welcome walkthrough, guardrails, and MCP wired in from first launch.
Read more →
04
Guardrails & MCP
Markdown rules and Mekari IAM–backed MCP tools keep assistants inside FSD, documented APIs, and platform security boundaries.
Read more →
05
Extension catalog
Install the core agent, AI context hub, MCP explorer, and seven guardrail packs — standalone VSIX or bundled in the branded IDE.
Read more →

01macOS bridge

The local runtime for Officeless coding

Every Code Intelligence editor surface talks to a single HTTP API on 127.0.0.1:8765. The macOS menu bar app hosts that bridge — provider routing, privacy mode, Sparkle updates, and request transparency without a Dock icon.

Apple Intelligence — on-device Foundation Models with optional Private Cloud Compute on supported Macs.
Alibaba Token Plan — organization-hosted Qwen, DeepSeek, Kimi, and related models through an Anthropic-compatible API.
Ollama & BYOK — point the bridge at self-hosted or cloud endpoints when Apple Intelligence is not required.

Download macOS app See architecture

Officeless Code Intelligence macOS menu bar popover showing Apple Intelligence ready and localhost bridge on port 8765 — Menu bar control plane

02VS Code extension

Copilot-style coding inside VS Code

Open the Code Intelligence panel from the activity bar. Describe what to build, ask about your open project, and get multi-turn answers powered by the local macOS bridge — no cloud required in local-only mode.

Workspace-aware context — scans your open VS Code folder, includes file lists and contents, and uses the active editor path (not generic placeholders).
Auto-apply code — fenced code blocks from assistant replies can be written straight into workspace files, Copilot-style. Apply individual blocks manually when you prefer review first.
Include open file & selection — toggle file context, attach the current editor selection, and pick general or devops chat profiles.
Whole-file inline completion — set contextLines: 0 to send the entire open file as completion context.

Install VS Code extension More screenshots

VS Code sidebar chat auto-applying generated TypeScript code to main.ts — Live in VS Code

03Officeless Code Intelligence IDE

A branded editor built for Officeless teams

A branded editor with embedded guardrails, platform context, design system rules, and bring-your-own-key AI — so assistants stay inside FSD, APIs, and Mekari standards.

Not a separate runtime — VS Code OSS with Officeless branding and safe defaults
macOS universal DMG and Windows EXE/ZIP distribution
Officeless MCP + seven guardrail extensions compiled in

Download Code Intelligence IDE Extension catalog & setup

Officeless Code Intelligence IDE — VS Code with embedded guardrails and MCP

04Guardrails & MCP

Keep AI inside Officeless boundaries

Markdown rules bundled with Officeless Code Intelligence IDE constrain assistants so output stays aligned with FSD, documented APIs, and Mekari security standards — plus Officeless MCP for workflow and data tools.

In the branded IDE: Command Palette → Officeless: Apply Guardrails to This Project writes .officeless-ai/ into your workspace.

Embedded guardrails

Seven built-in extension modules mirror the ai-context/ library: global rules, platform APIs, frontend, design system, prompts, roles, and BYOK models.

Welcome & walkthrough

Officeless: Open Welcome Guide, step-by-step walkthrough on the Welcome page, and per-module guardrail browsers from the Command Palette.

BYOK-ready AI

Configure OpenAI, Claude, Bedrock, Cursor, Copilot, or self-hosted models. Built-in Microsoft Copilot stays optional to avoid forced sign-in.

Project sync

Copy rules into any repo as .officeless-ai/, Cursor rules, Continue config, and Copilot instructions — from the IDE or npm run sync-context.

Service-layer discipline

AI is steered toward src/services/officeless/ for HTTP, Storybook for UI, and human-reviewed tests before handover.

Officeless MCP

Mekari IAM OAuth connects the IDE to the Officeless MCP gateway — workflow, forms, data manager tools synced for Copilot Chat.

Module	Focus
Global	System context, AI boundaries, security, dev/test/release rules
Platform	Officeless API integration, auth, workflows, embedded UI
Design System	Storybook — components and theming; no random UI libraries
Frontend	React, TypeScript, services, state, error handling
Prompts	Generate, review, refactor, and validate templates
Roles	Architect, QA, and solution engineer guides
AI Models (BYOK)	Per-provider BYOK notes (Cursor, Claude, OpenAI, Copilot, Bedrock, self-hosted)

Allowed

Officeless custom frontend and dashboards
API service wrappers under src/services/officeless/
Design-system UI, tests, Storybook

Not allowed

Custom auth replacement or direct database access
Undocumented APIs, production secrets, real customer data
Workflow bypass or unapproved UI libraries

Full guides: code-intelligence.vas.mekari.com/guides · Code Intelligence IDE download

05Extension catalog

VSIX packs for every Officeless coding surface

Install the core agent, AI context hub, MCP explorer, and guardrail modules individually — or use the branded IDE where they ship preloaded.

Standalone VS Code

Code Intelligence

mekari-vas.officeless-code-intelligence

Copilot-style coding chat and inline completion — Apple Intelligence (macOS menu bar app), Claude via Code Intelligence Gateway, Alibaba Cloud Model Studio Token Plan, or ai.suherman.net (Ollama).

Extensions → ⋯ → Install from VSIX… — or use Officeless: Check for Extension Updates when already installed.

Branded IDE + VSIX

Officeless AI Context

mekari-vas.officeless-ai-context

Hub extension — welcome walkthrough, guardrail catalog, connect AI providers, copy Cursor rules.

Built into Code Intelligence IDE; install VSIX on standard VS Code for guardrail hub commands.

Branded IDE + VSIX

Officeless MCP

mekari-vas.officeless-mcp

Mekari IAM OAuth → Officeless MCP gateway — tools explorer and native MCP sync for Copilot Chat.

Command Palette → Officeless: Connect MCP after install.

Branded IDE + VSIX

Guardrails (all modules)

mekari-vas.officeless-guardrails

All Officeless guardrail modules in one extension — global, platform, frontend, design system, prompts, roles, and AI models.

Command Palette → Officeless: Browse All Guardrails — single VSIX for all guardrail modules.

Recommended Marketplace extensions

Optional — install via Extensions view or npm run install:extensions from officeless-code-intelligence.

Extension	Extension ID	Purpose
GitHub Copilot	`github.copilot`	Inline suggestions and chat (sign in with org subscription).Extensions view → search by name, or npm run install:extensions from the repo.
GitHub Copilot Chat	`github.copilot-chat`	Copilot chat panel in VS Code.Extensions view → search by name, or npm run install:extensions from the repo.
Continue	`continue.continue`	BYOK chat and autocomplete — OpenAI, Anthropic, Ollama, and more.Extensions view → search by name, or npm run install:extensions from the repo.
Claude Code	`anthropic.claude-code`	Anthropic Claude assistant for VS Code.Extensions view → search by name, or npm run install:extensions from the repo.
Gemini Code Assist	`google.geminicodeassist`	Google Gemini coding assistant.Extensions view → search by name, or npm run install:extensions from the repo.
Playwright Test	`ms-playwright.playwright`	End-to-end testing for web apps.Extensions view → search by name, or npm run install:extensions from the repo.
ESLint	`dbaeumer.vscode-eslint`	JavaScript and TypeScript linting.Extensions view → search by name, or npm run install:extensions from the repo.
Prettier	`esbenp.prettier-vscode`	Code formatter for JS/TS, JSON, Markdown, and more.Extensions view → search by name, or npm run install:extensions from the repo.
Tailwind CSS IntelliSense	`bradlc.vscode-tailwindcss`	Autocomplete and linting for Tailwind classes.Extensions view → search by name, or npm run install:extensions from the repo.
Code Spell Checker	`streetsidesoftware.code-spell-checker`	Spelling checker for source code and docs.Extensions view → search by name, or npm run install:extensions from the repo.
YAML	`redhat.vscode-yaml`	YAML language support and validation.Extensions view → search by name, or npm run install:extensions from the repo.
REST Client	`humao.rest-client`	Send HTTP requests from `.http` files.Extensions view → search by name, or npm run install:extensions from the repo.

In action

Screenshots from VS Code

Sidebar chat, auto-apply, and extension settings — connected through the local macOS bridge when using Apple Intelligence.

Copilot-style chat with auto-apply

Ask in plain language from the Code Intelligence activity bar panel. The assistant connects to Apple Foundation Models through the local bridge, respects your open workspace, and can auto-apply fenced code blocks straight into files — here a TypeScript service was created in main.ts after “write code for hello world”. Toggle Include open file, Auto apply, and DevOps profile from the toolbar.

VS Code Code Intelligence sidebar chat panel with auto-apply creating main.ts from a hello world prompt — VS Code · sidebar chat

VS Code Officeless Code Intelligence extension settings for context lines, auto apply, and workspace files — VS Code · settings

Configure context for your whole codebase

Fine-tune the extension in Settings: point Api Base Url at the local bridge, set Context Lines to 0 for the entire open file (or any large value for whole-file inline completion), enable Auto Apply Code and Include Workspace Files so chat sees your real project paths, and switch Chat Profile between general coding and DevOps Central.

Capabilities

Everything the platform includes

Beyond sidebar chat and the macOS bridge — terminal REPL, DevOps profiles, selection commands, and flexible provider routing for Officeless workflows.

Sidebar chat panel

Copilot-style chat in the VS Code activity bar. Multi-turn conversations with workspace file context, open-file awareness, and optional auto-apply of code blocks.

Auto-apply to workspace

Assistant replies with fenced code can be written directly into project files — create or update paths inferred from your prompt and the open editor.

Whole-file inline completion

Set contextLines to 0 to send the entire open file to the bridge. Ghost-text completions use prefix, suffix, and full nearbyCode for better suggestions.

Terminal chat (Codex CLI style)

Prompt-only REPL in Terminal.app or VS Code's integrated terminal. Attach files, switch DevOps profile, and keep multi-turn history in the session.

Workspace-aware prompts

Chat and CLI use your open VS Code folder as the project root — real paths, file lists, and contents instead of generic placeholders.

DevOps Central profile

Tuned for K3s, Argo CD, Terraform, secret-broker, vas-access-hub, and hetzner-platform-foundation GitOps workflows.

Explain, refactor, and ask

Select code and run Explain Selection, Refactor Selection, or Ask About Selection — context goes to the local bridge and opens beside your editor.

Apple Foundation Models

On-device language model through Foundation Models on macOS 26+, with Private Cloud Compute where available. Provider routing to Ollama, OpenAI, or Anthropic when configured.

Menu bar control plane

Lightweight macOS agent with the Code Intelligence logo in the menu bar. Shows provider status, privacy mode, Sparkle auto-update, and localhost :8765 health.

Terminal chat

Codex-style coding in your terminal — powered by Apple Intelligence

Not a code editor. Like OpenAI Codex or Gemini CLI: open a terminal, type a prompt, get coding answers. All inference stays on your Mac through the menu bar bridge — lighter on memory than running a full IDE AI stack.

Prompt-only REPL — no file edits, no agent sandbox. Just chat for coding questions.
DevOps Central profile — tuned for K3s, Argo CD, Terraform, secret-broker, and platform GitOps on Hetzner.
Attach context — pipe a file, use /file Makefile, or ask from VS Code with selected code.
Multi-turn history — conversation is kept in the session and sent to POST /v1/chat.

Install and try it See the API

Quick start

# 1. Launch the macOS menu bar app (bridge on :8765)

# 2. Install CLI from the repo
git clone https://bitbucket.org/jojocoders/officeless-intelligence-runtime.git
cd officeless-intelligence-runtime
npm run cli:install

# 3. Start interactive chat
npm run chat

# DevOps Central / platform engineering profile
npm run chat -- --devops

# One-shot question
npm run chat -- -m "How do I deploy secret-broker to prod-vas?"

# Attach a file as context
npm run chat -- --devops -f gitops/applications/secret-broker/deployment.yaml \
  -m "Explain this deployment and common failure modes"

REPL commands

/help — list commands
/clear — reset conversation
/status — bridge provider info
/profile devops — switch to DevOps profile
/file path — load file into next prompt
/exit — quit

VS Code integration

Prefer a panel over the terminal? Click the Code Intelligence icon in the VS Code activity bar for Copilot-style sidebar chat with workspace context and auto-apply. Or use Command Palette (⇧⌘P) → Officeless: Open Chat Panel.

For terminal REPL: Officeless: Open Terminal Chat opens Officeless Chat in the integrated terminal. Uses your workspace folder and chatProfile from VS Code settings. See the VS Code install guide for sidebar chat, inline completion, and settings.

Architecture

One local API, many editor surfaces

VS Code uses one extension with two inference backends. Point at the local macOS bridge for Apple Intelligence, or at Alibaba Token Plan for organization-hosted models — same chat UI, inline completion, and auto-apply in both modes.

├─VS Code sidebar chat · inline completion · model picker

├─Choose backend: Apple Intelligence or Alibaba Token Plan

├─Apple Intelligence → macOS bridge → Foundation Models

├─Token Plan → Anthropic-compatible API → Qwen / DeepSeek / Kimi

└─Workspace context · file scan · code apply

Inline completion

POST /v1/code/complete
{
  "language": "typescript",
  "filePath": "src/api/users.ts",
  "prefix": "... entire file before cursor when contextLines is 0",
  "suffix": "... after cursor",
  "nearbyCode": "... full open file",
  "instruction": "complete the next tokens at the cursor"
}

Sidebar chat

POST /v1/chat
{
  "message": "Add a health check route to server.ts",
  "filePath": "/Users/you/project/src/server.ts",
  "context": "... workspace file list + open file + selection",
  "profile": "general",
  "history": [
    { "role": "user", "content": "..." },
    { "role": "assistant", "content": "..." }
  ]
}

Endpoints

/v1/status
/v1/code/complete
/v1/code/explain
/v1/code/refactor
/v1/chat

Privacy

Your code should not leave your machine by surprise

Officeless Code Intelligence is designed for teams that want Apple Intelligence in the editor without routing every keystroke through a remote SaaS API.

Local-only by default

Code context stays on your Mac. Cloud providers are disabled until you change privacy mode.

Optional path redaction

Strip full file paths from prompts before they reach any model backend.

Local request log

Recent requests are stored in the menu bar app for transparency, not uploaded anywhere.

Localhost binding

The bridge listens on 127.0.0.1 only. Nothing is exposed to your network.

Get started

Choose your setup path

Install the VS Code extension, then choose Apple Intelligence (local macOS bridge) or Alibaba Token Plan (cloud). See the VS Code install guide for both setup paths.

Code Intelligence IDE

Branded VS Code with embedded guardrails, Officeless MCP, welcome walkthrough, and BYOK-ready AI — macOS and Windows installers.

Download branded IDE →

VS Code

One extension — configure Apple Intelligence or Alibaba Token Plan. Sidebar chat, model picker, auto-apply, and inline completion.

Install extension →

macOS app

Required for Apple Intelligence backend. Menu bar bridge on localhost :8765, Sparkle auto-update, on-device Foundation Models.

Download & install →

Terminal chat

Codex-style REPL — requires the macOS bridge (Apple Intelligence backend). DevOps profile and file attachments.

Set up terminal chat →