Documentation · Officeless platform

User guides & guardrails

Step-by-step setup for Officeless Insider — every Code Intelligence plugin is pre-installed and locked. Connect MCP, sync rules into your repo, and configure BYOK AI.

Overview

Officeless Insider is branded VS Code Insiders with every Code Intelligence extension pre-installed. Plugins cannot be removed but update when new VSIX builds ship. The IDE teaches assistants how to work on Officeless projects; your repo still holds the implementation (React UI, services, tests).

Four layers to remember
  • Platform — workflows, auth, APIs (Officeless)
  • Your project — UI, src/services/officeless/, tests
  • Guardrails — markdown rules the IDE (and synced tools) load for AI
  • Officeless MCP — live platform tools (projects, forms, workflows) your AI can call securely

Code Intelligence plugins (publisher: mekari-vas)

Extension IDs use publisher mekari-vas, aligned with the *.vas.mekari.com platform domain. Insider embeds the latest stable VSIX at build time; updates ship via Officeless: Check for Extension Updates or from code-intelligence.vas.mekari.com/download/vscode.

Extension ID Install on standard VS Code
mekari-vas.officeless-code-intelligence Download VSIX → Extensions → ⋯ → Install from VSIX…
mekari-vas.officeless-ai-context Hub extension — required for guardrail browser commands
mekari-vas.officeless-mcp Then Command Palette → Officeless: Connect MCP
mekari-vas.officeless-guardrails-global-cimb Optional per-module VSIX; Insider includes all eight built-in

macOS menu bar bridge bundle ID: com.mekari.vas.officeless-code-intelligence · Insiders IDE: com.mekari.vas.officeless.ai.insiders

1. Install the IDE

macOS

  1. Download the latest .dmg from the home page.
  2. Open the disk image and drag Officeless AI Insiders to Applications, then eject the volume.
  3. Launch from Applications. Releases are notarized by Apple — no Terminal workaround required.

Windows

  1. Download the Setup EXE (recommended) or portable ZIP from the home page.
  2. Setup EXE: run the installer and launch from the Start menu.
  3. ZIP: extract the archive and run Officeless AI Insiders.exe.

2. First launch & in-IDE walkthrough

  1. Launch Officeless AI Insiders from Applications.
  2. Open the welcome walkthrough: Help → Welcome → choose Officeless AI Insiders.
  3. Open the full guide anytime: Cmd+Shift+POfficeless: Open Welcome Guide.
  4. Browse all guardrail packs: Officeless: Browse All Guardrail Extensions.

Built-in Microsoft Copilot is disabled in this build. Configure a real AI provider next — do not use the fake Continue with Officeless sign-in screen.

3. AI chat & inline completions

Not a real Officeless login

If the IDE shows Sign in to use GitHub Copilot with Continue with Officeless, close it. This build has no Officeless cloud AI account. Use one of the options below.

Option A — Continue (BYOK, recommended)

  1. In Insiders: Extensions → install Continue (continue.continue).
  2. In the guardrail kit repo:
    npm run setup:byok
    npm run sync-context -- --target /path/to/your-project --continue --all-tools
  3. Open the Continue side bar → add a model (OpenAI, Anthropic, Ollama, etc.) using keys from .env.local (never commit keys).
  4. Use Continue for chat and edits — not the generic Copilot sign-in panel.

Option B — GitHub Copilot (Marketplace)

  1. Extensions → install GitHub Copilot and GitHub Copilot Chat (publisher: GitHub).
  2. Command PaletteGitHub Copilot: Sign In and authenticate with your GitHub account (per org policy).
  3. Sync instructions:
    npm run sync-context -- --target /path/to/your-project --copilot

Option C — Cursor or local AI

  • Cursor: sync guardrails and use Cursor IDE, or run Officeless: Copy Cursor Rules to Workspace in Insiders.
  • Local (Ollama / Open WebUI): use Continue with an OpenAI-compatible URL (e.g. http://localhost:11434/v1).

In the IDE: Cmd+Shift+POfficeless: Open AI Setup Guide for the full in-app guide.

4. Open your Officeless project

  1. File → Open Folder and select your implementation repo (dashboard, portal, or embedded UI).
  2. Ensure project context exists where possible: project-context/fsd/ (authoritative scope), project-context/api-contracts/, and src/services/officeless/.
  3. Optional: use the kit workspace template — vscode/workspace-template/officeless.code-workspace — after cloning the guardrail kit.

5. Officeless MCP

Officeless MCP connects the IDE to the Officeless platform through the MCP gateway at https://mcp-officeless.mekari.com/mcp. After you connect, AI assistants (Copilot Chat, Continue, Claude, Gemini, and others) can call platform tools — list Studio projects, inspect forms and tables, search knowledge, run reports — using real data instead of guessing.

Guardrails teach the model how to build on Officeless; MCP gives it what exists on the platform. Use both together: browse guardrail modules for coding rules, then ask the chat to invoke MCP tools for live project metadata.

Connect to the MCP gateway

  1. Cmd+Shift+POfficeless: Connect MCP (or open the MCP Connect side bar).
  2. Enter your organisation’s OAuth client ID and client secret. Secrets are stored in VS Code SecretStorage only — never in settings.json or your repo.
  3. Click Save credentials, then Connect. When connected, the status badge shows Connected and the gateway reports how many tools are available (typically 60+).
  4. The extension syncs into VS Code’s native MCP config so GitHub Copilot Chat can call Officeless tools directly — no manual bearer token paste in chat.
Officeless MCP Connect panel, tools catalog with Studio tools, and Copilot Chat invoking studio_list_projects
Left: MCP Connect — endpoint, OAuth credentials, connection status. Center: Tools catalog — browse Studio, Reports, Knowledge, and other tool groups; use Ask in chat or Run in Explorer on any tool. Right: Copilot Chat reads the credential manifest, then calls studio_list_projects without exposing secrets in the prompt.

Browse tools & run from the IDE

After login, open the live tool catalog:

  • Officeless: Open MCP Tools Catalog — searchable list of all gateway tools, grouped by area (Studio, Reports, Dashboards, Knowledge, Connections, …).
  • Officeless: Open MCP Explorer — run a tool with parameters and inspect the JSON response in the IDE.
  • MCP: List Servers (VS Code) — confirm officeless-mcp is connected for Copilot Chat.

Common Studio tools include:

Tool Purpose
studio_list_projects Paginated list of Officeless Studio projects (IDs, names, metadata)
studio_list_applications Applications within a project
studio_list_pages / studio_list_form Pages and forms in Studio
studio_list_tables / studio_list_workflows Data tables and workflow definitions

Each tool card offers Ask in chat (pre-fills a safe prompt with credential guidance), Copy name, and Schema for parameters.

Use MCP in AI chat

Example workflow — retrieve projects with pagination:

  1. Connect MCP and configure your AI provider (Continue or Copilot — see section 3).
  2. Ask in chat, e.g. “List Officeless Studio projects with pagination.”
  3. The assistant loads the credential manifest from extension storage (path shown in chat — for the AI only, not for pasting secrets) and invokes studio_list_projects.
  4. Review the returned project list; use guardrails when you implement UI or services against that data.
Officeless Guardrails catalog, document preview, and chat using MCP to list projects
Guardrails (left) and MCP-powered chat (right) side by side: browse AI Models (BYOK) or other modules while the assistant calls platform tools. The center preview shows per-module docs; Command Palette → Officeless: Browse … Guardrails opens any file.
Security
  • OAuth client secret and access tokens live in SecretStorage only.
  • Never paste client secrets or bearer tokens into chat or commit them to git.
  • The generated credentials.md manifest (under extension global storage) tells the AI where credentials are — not the secret values.
  • Synced repos may include .officeless-ai/mcp/credential-locations.md as a static reference after guardrail sync.

Command Palette — MCP

Officeless: Connect MCP

Open connect panel; save OAuth credentials

Officeless: Open MCP Tools Catalog

Browse all gateway tools by category

Officeless: Open MCP Explorer

Run tools and inspect responses

Officeless: Disconnect MCP

Logout and clear stored tokens

Your organisation supplies the OAuth client ID (public). Client secret is issued per environment — ask your Officeless / Mekari platform admin if you do not have one.

6. Sync guardrails into your repository

Bundled rules live inside the IDE. To give Cursor, Continue, Copilot, and agents the same context inside your repo, sync into the project (one-time per project, re-run when guardrails update).

Recommended: apply from Officeless AI Insiders (in-IDE)

  1. Open your project folder in Officeless AI Insiders.
  2. Cmd+Shift+POfficeless: Apply Guardrails to This Project.
  3. Select which tool integrations you want (Cursor / Copilot / Continue).

This creates (in your repo):

Path Purpose
.officeless-ai/ Full guardrail context synced into the repository
project-context/ Scaffold folders: fsd, api-contracts, qa, brd, ai
.cursor/rules/officeless-ai.mdc Cursor rules (optional)
.github/copilot-instructions.md GitHub Copilot instructions (optional)
.continue/config.yaml Continue config wired to .officeless-ai/ (optional)
.officeless-ai/mcp/ MCP credential reference and usage notes for AI chat (optional)

Alternative: sync from the guardrail kit (CLI)

git clone https://bitbucket.org/jojocoders/officeless-ai-vscode-guardrail-kit.git
cd officeless-ai-vscode-guardrail-kit
npm install
npm run sync-context -- --target /path/to/your-officeless-project --all-tools

Tip: if you only need Cursor rules (without copying full context), run Officeless: Copy Cursor Rules to Workspace.

7. Configure AI (bring your own key)

  1. In the guardrail kit repo:
    npm run setup:byok
    Follow prompts; output is .env.local (gitignored).
  2. Install recommended extensions (if not already in the Insiders profile):
    npm run install:extensions
  3. Use Continue (or your provider) inside Insiders with Officeless context. Per-provider notes are under guardrail module Models (BYOK).

Never commit API keys. Use env vars or your org’s secret store.

8. Use guardrails in daily development

Guardrails are markdown rules — use them actively, not only at install time.

Workflow (recommended)

  1. Start from FSD — functional spec in project-context/fsd/ is authoritative; BRD is background only.
  2. Load boundaries before coding — Command Palette → Officeless: Open AI Boundaries or Open System Context.
  3. UI work — open Officeless: Open Design System Rules and use Storybook; no ad-hoc UI libraries.
  4. API work — all HTTP via src/services/officeless/; open Officeless: Open API Integration Rules.
  5. Review output — human review + tests; no hallucinated endpoints or fields.

Command Palette reference

Officeless: Open Welcome Guide

Full IDE overview and links

Officeless: Open Welcome Walkthrough

Step-by-step on Welcome page

Officeless: Browse All Guardrail Extensions

Quick-pick every module

Officeless: Browse … Guardrails

One command per module (Global, Platform, …)

Officeless: Open System Context

Global platform + scope rules

Officeless: Open AI Boundaries

Allowed / not allowed for AI

Officeless: Copy Cursor Rules to Workspace

Sync Cursor rules into open repo

9. Guardrail modules — when to use each

Module Use when you…
Global Need system context, security, dev/test/release rules, or AI boundaries
Platform Integrate Officeless APIs, auth, workflows, or embedded UI
Design system Build UI components, theming, or Storybook usage
Frontend Work on React, TypeScript, services, state, or errors
Prompts Run structured tasks: generate, review, refactor, validate scope
Roles Adopt architect, QA, solution engineer, or advisor personas
Models (BYOK) Configure OpenAI, Claude, Gemini, Cursor, Copilot, or self-hosted
CIMB advisor Assess Officeless Gen1 (CIMB UAT) — forms, tables, connectivity

Open any module via Officeless: Browse [module] Guardrails in the Command Palette.

10. Prompt templates

Under ai-context/prompts/ (and synced .officeless-ai/prompts/):

  • generate-feature — scaffold from FSD with guardrails applied
  • review-code — check against boundaries and service layer
  • refactor — safe refactors within Officeless patterns
  • validate-scope — confirm work stays inside FSD
  • assess-officeless-gen1 — Gen1 assessment (with CIMB pack)

Paste or reference these at the start of an AI session so the model inherits the same constraints as the IDE. For smaller local models, load only 00-system-context.md, 02-ai-boundaries.md, and the relevant FSD section.

11. Per AI tool

Tool How guardrails load
Insiders (built-in) Bundled extensions + Command Palette commands
Cursor .cursor/rules/officeless-ai.mdc after sync
Continue .continue/config.yaml docs providers
GitHub Copilot .github/copilot-instructions.md; native MCP via officeless-mcp when connected
Officeless MCP Built-in extension; Copilot Chat + Explorer tools after Officeless: Connect MCP
Cline custom-instructions.md in extensions template

Conflict resolution: project FSD wins on business scope; global guardrails win on security, design system, and service architecture; API contracts win on field names.

12. Verify & troubleshoot

From the guardrail kit repo:

npm run doctor
npm run validate:context
  • Blank window / “default chat agent” — re-run npm run build:vscode:prepare in the kit (dev builds).
  • AI ignores rules — confirm .officeless-ai/ exists and you referenced boundaries in the prompt.
  • Wrong API shape — point the model at project-context/api-contracts/.
  • MCP not connected — run Officeless: Connect MCP; check MCP: List Servers shows officeless-mcp connected.
  • Copilot cannot call MCP tools — confirm Copilot Chat is signed in and native MCP sync is enabled (officeless.mcp.syncNativeMcpConfig is not false).
  • Chat exposes secrets — stop the session; credentials must stay in SecretStorage. Reload credential guidance via Officeless: Open MCP Tools Catalog.

← Back to home · Download latest build