ChipOS ChipOS Docs
Back to ChipOS
OFFICIAL DOCUMENTATION

ChipOS Documentation

The complete guide to the AI Operating System for chip development. Start from login, pick a workspace and mode, then use ChipOS to manage knowledge, projects, and AI workflows end-to-end.

First Launch

Login, choose a mode, and get started fast

Modes & Packs

Agent, Chat, and Zendesk workflows

Knowledge Base

Index docs, repos, and code examples

What is ChipOS?

ChipOS is the AI Operating System for chip development — think macOS for chips, with AI woven into design, verification, and knowledge. It combines a dedicated desktop environment, an embedded dashboard, and a cloud or local backend so you can build, search, and automate with AI across your entire workflow.

3
Modes: Agent, Chat, Zendesk
RAG
Knowledge + repo search
MCP
Connect any AI assistant

For Chip Teams

  • Agent packs for RTL verification, physical design, and CPU development
  • Knowledge base with repo ingestion and code-aware search
  • Projects and tasks to track specs, implementation, and QA

For Support & Ops

  • Zendesk Mode for ticket analysis and AI-drafted responses
  • ChipChat fullscreen Chat Mode with provider choice
  • Hosted or local backend options for privacy needs

Core Flow

ChipOS is designed as a simple, repeatable loop: sign in, pick a workspace, choose your mode, configure providers, add knowledge, and connect MCP for your assistants.

1 Sign In 2 Workspace 3 Mode 4 Dashboard 5 Knowledge 6 MCP

Chat Mode always uses the hosted MCP endpoint, even if your backend is set to local.

Chat Mode Path

  1. Pick Chat Mode in the workspace selector.
  2. ChipOS SSO launches ChipChat fullscreen.
  3. Select a model and chat with MCP-powered context.

Agent Mode Path

  1. Choose Agent Mode and a pack for your domain.
  2. Open the dashboard to configure providers and knowledge.
  3. Connect MCP to Claude Code, Cursor, or Windsurf.

Login & Authentication

Unified sign-in. Use email/password or OAuth (Google + GitHub on desktop, Google + Microsoft on web/chat). Admins can require login, and local mode can skip sign-in if enabled.

ChipOS uses secure authentication with a dedicated sign-in panel when required and stores credentials securely. Sessions refresh automatically so you can reopen ChipOS without signing in each time.

Google

One-click sign-in

GitHub

For developers

Email + Password

Standard login with recovery options

Microsoft

Web dashboard + Chat

Installation

Standalone by design

ChipOS is a standalone application with its own data folder. It won't interfere with your existing editors or IDEs.

1

Download the DMG

Apple Silicon (ARM64) Intel (x64)

macOS 10.15+ supported (11+ recommended)

2

Open the DMG and drag to Applications

The app is notarized by Apple — no security warnings.

3

Launch ChipOS

Find it in your Applications folder or Spotlight.

First Launch Flow

From the login screen to your first project in a few minutes.

1

Sign In (If Required)

If login is required, a secure sign-in panel appears (non-blocking). Use email/password, Google, or GitHub. In local mode, admins can allow sign-in to be skipped.

ChipOS → Sign In → Continue

2

Choose Workspace + Mode

Pick a folder or .code-workspace, then select your startup mode. Agent Mode also asks for an agent pack.

Agent Mode

Chat Mode

Zendesk Mode

Agent Packs

3

Launch Your Mode

Agent Mode opens the ChipOS sidebar and (if enabled) the dashboard. Chat Mode launches ChipChat fullscreen with the hosted MCP endpoint and default model.

Agent Mode

Controls + Dashboard + Claude commands

Chat Mode

ChipChat fullscreen + MCP context

Connect MCP + Start a Command

Copy the MCP config from Controls and add it to Claude Code, Cursor, or Windsurf. Chat Mode uses hosted MCP by default. Then try your first command:

/init-chipos Setup this workspace for ChipOS tracking

ChipOS will scaffold tracking and connect your workspace to the knowledge base.

Configuration

ChipOS settings can be configured through the Settings panel or settings.json.

Setting Default Description
chipos.startupMode "agent" Startup mode: agent, chat, or zendesk
chipos.agentMode "spec-developer" Selected agent pack for Agent Mode
chipos.backend.mode "hosted" Backend mode: hosted (cloud) or local (Docker)
chipos.backend.hostedUrl https://api.futureatoms.com Hosted backend API URL
chipos.backend.apiUrl https://api.futureatoms.com Local API override (set to http://localhost:8181 in local mode)
chipos.backend.supabaseUrl "" Hosted auth project URL
chipos.backend.projectRef "" Hosted auth project ref (Settings → General)
chipos.auth.requireLogin true Require login on startup
chipos.auth.skipLoginForLocalMode true Skip login when running locally
chipos.showWelcomeOnStartup true Open dashboard on startup
AGENT PACKS

Agent Packs & Contexts

Choose an agent pack in the workspace selector. Packs install Claude Code command sets and tailor prompts for your hardware or software domain.

CORE Core Packs

Spec Developer

Turn specs into structured tasks, docs, and checklists.

Documentation

Generate and maintain project documentation with AI assistance.

Web Development

UI and tooling support for internal dashboards and apps.

RTL & PHYSICAL RTL & Physical Packs

RTL Verification

SystemVerilog + UVM workflows and testbench support.

RTL Debug

Waveform analysis, assertions, and failure triage.

Physical Design

Place-and-route assistance and physical flow checks.

Physical Design Verification

DRC/LVS verification and sign-off readiness.

SYSTEMS Systems & SDK Packs

Hardware Programmer

Low-level programming and firmware workflows.

CPU Development

RISC-V architecture planning and subsystem design.

Raspberry Pi

Embedded workflows for Pi-based prototyping.

Axelera Voyager SDK

AI acceleration pipelines and Metis tooling.

STARTUP MODES

Agent, Chat & Zendesk

Choose a startup mode in the workspace selector. Each mode opens a different surface and configures ChipOS for your workflow.

Agent Mode

IDE-first workflow

Use ChipOS inside the IDE with a selected agent pack. Agent Mode installs Claude Code commands, opens the dashboard, and connects MCP tools for knowledge, projects, and tasks.

  • Installs the selected agent pack to .claude/commands
  • Dashboard access for Knowledge, Projects, MCP, and Settings
  • Works with Claude Code, Cursor, and Windsurf via MCP
  • Supports Agent Work Orders and repo ingestion workflows
  • Optional local backend for private deployments

Chat Mode

ChipChat fullscreen

Chat Mode opens ChipChat in a fullscreen webview. ChipOS SSO signs you in, then the chat UI loads with the hosted MCP endpoint and your default model.

  • ChipOS SSO for ChipChat
  • Multi-provider model selection (OpenAI, Anthropic, Google, OpenRouter, Grok)
  • Hosted MCP endpoint for Chat Mode (even if backend is local)
  • Default model: zai-org/GLM-4.7:fastest
  • MCP tools available for knowledge and project context
  • Press Cmd+Shift+F / Ctrl+Shift+F to toggle fullscreen, Esc to exit
  • Chip AI extension is available but disabled by default

Zendesk Mode

For Support Teams

LLM-powered support workflows for Zendesk. Query your portal with AI assistance, draft responses, and update tickets directly from Chat or Agent Mode.

  • Authenticate with admin or personal Zendesk accounts
  • Search tickets, comments, and Help Center articles
  • Draft and post replies using any configured LLM
  • Connected via the Zendesk MCP server
Zendesk setup

Provide your Zendesk subdomain, email, and API token in secure settings to enable ticket tools.

Dashboard & Navigation

The ChipOS Dashboard (ACF UI) is your control center. Open it from Controls or let it launch on startup. It hosts onboarding, knowledge, projects, MCP settings, and provider configuration.

Welcome & Onboarding

Verify API keys, provider setup, and onboarding status.

Knowledge & Projects

Add sources, browse knowledge, and manage project tasks.

MCP & Providers

Copy MCP configs and monitor provider connectivity.

Settings

Configure hosted vs local mode, RAG, and Ollama.

Local dev fallback

In local mode, the dashboard tries the dev server first and falls back to the packaged UI if it is unavailable.

AI MODELS

AI Providers & Models

Multi-Provider Routing

Bring your own keys and route per task

ChipOS supports multiple LLM providers and local runtimes. Configure keys once in Settings, then select models per workflow or let ChipOS route requests automatically.

OpenAI Anthropic Google Ollama

Provider Categories

Anthropic

  • • Claude family models
  • • Great for reasoning and specs

OpenAI

  • • GPT-4 class models
  • • Strong for code + tool use

Google

  • • Gemini models
  • • Fast multimodal workflows

Aggregators / Gateways

  • • OpenRouter and other gateways
  • • One key, multiple vendors

Local / On-Prem

  • • Ollama, LMStudio, Llama.cpp
  • • Keep models on your network

Custom

  • • Custom endpoints and gateways
  • • Enterprise routing rules

Provider Status & Routing

The Settings dashboard shows live connectivity for OpenAI, Anthropic, Google, OpenRouter, Grok, and Ollama. Use health checks to verify keys and pick the best model for specs, verification, or support automation.

Pro Tip: Use health checks to verify your API keys are working, then switch models per task in Agent Mode based on your needs.

SECURITY & COMPLIANCE

Enterprise-Grade Security

ChipOS is built with security-first principles. OWASP LLM Top 10 (2025) and OWASP API Top 10 (2023) aligned, NIST AI RMF documented, with Trivy + Gitleaks security scans in CI.

NIST AI RMF

Documented & Tracked

Documented alignment with the NIST AI Risk Management Framework, including governance, measurement, and risk tracking.

AI Risk Assessment Model Governance

OWASP API Top 10

2023 Aligned

Hardened API surface with strict validation, rate limits, and secure defaults across hosted and local deployments.

Input Validation Rate Limiting

OWASP LLM Top 10

2025 Aligned

Protections against prompt injection, data leakage, and model misuse with automated security tests.

Trivy Scans Gitleaks Scans

Privacy First

No Telemetry by Default

Telemetry is disabled by default. Secrets are stored securely and row-level security isolates tenant data in hosted mode.

Zero Telemetry RLS Enabled

Knowledge Base

The Knowledge Base indexes docs, repos, and uploads so ChipOS can answer questions with code-aware, semantic search. Use it for specs, RTL, SDK docs, and internal playbooks.

Web Crawling

Crawl documentation sites with sitemap and llms.txt discovery.

  • Allow/deny path filters
  • Progress tracking with chunk counts
  • HTML to markdown conversion

Repository Ingestion

Index GitHub repositories with code-aware chunking and metadata.

  • Code chunking with line ranges
  • Language detection and metadata
  • Stores snippets as code examples
  • Use “Repository (Best for Code)” in Add Knowledge

Document Upload

Upload and index PDFs, DOCX, Markdown, and text files.

  • PDF text extraction
  • DOCX and HTML support
  • Automatic chunking + embeddings

Knowledge Inspector

Browse chunks and code examples with pagination.

  • View source-level metadata
  • Load-more pagination for large sources
  • Inspect code snippets in context

Projects & Tasks

Projects organize your work with tasks, documents, and linked knowledge sources. Tasks can be tracked in a Kanban board or table view with clear status and assignee fields.

Projects

  • Create projects with descriptions and GitHub links
  • Pin or edit projects from the dashboard
  • Link knowledge sources per project

Tasks

  • Statuses: todo, doing, review, done
  • Assign tasks to users or AI agents
  • Drag-and-drop Kanban or table view

Assistant Chat

The dashboard Assistant provides hardware-aware chat with syntax highlighting and citations. For fullscreen chat, use ChipChat from the workspace selector.

Supported Languages

Python SystemVerilog Verilog YAML JSON Bash
Chat surfaces

Chip AI is available but disabled by default. Chat Mode uses ChipChat as the primary fullscreen experience.

RAG-Connected

The chatbot automatically queries your knowledge base for hardware-specific questions, providing accurate, contextual answers.

MCP Server

The Model Context Protocol (MCP) server connects external assistants to your knowledge base, projects, and tasks. Choose between hosted mode for convenience or local mode for privacy.

Hosted + Local

Use hosted MCP or run locally for privacy

Tool Catalog

Knowledge, projects, tasks, and more

Secure Sessions

Auth-aware sessions with audited tool calls

Available MCP Tools

Tool Description
rag_search_knowledge_base Semantic search across indexed docs
rag_search_code_examples Find code snippets from indexed repositories
rag_get_available_sources List indexed knowledge sources
list_projects List projects and metadata
create_task Create tasks linked to a project

Claude Commands

When you open a workspace in Agent Mode, ChipOS installs 14 bundled Claude Code commands to .claude/commands/. These commands work across macOS and Linux builds.

Download Command Pack (zip)
/axe

Build one-shot pipelines with required output folders

/init-chipos

Initialize ChipOS project tracking in a workspace

/voyager-help

Query the ChipOS knowledge base with RAG

/agent

Axelera Voyager SDK workflow agent

/new-pipeline

Scaffold pipeline configs for models and tooling

/deploy

Deploy pipelines or models to target environments

/run

Run pipelines and inference tasks

/bench

Benchmark model performance metrics

/new-app

Create new application scaffolds

/debug

Debug pipeline, model, or inference issues

/test

Run tests for your workspace or pipelines

/build

Build artifacts and packages

/add-model

Add a custom model to the model zoo

/list-models

List and search available models

RAG OVERVIEW

RAG Overview

ChipOS blends hybrid search, reranking, and code-aware retrieval so answers cite the right docs and code for your workspace.

Hybrid Search + Reranking

Combines vector search with keyword signals, then reranks results for higher relevance.

Code-Aware Retrieval

Repository ingestion stores code examples with metadata so answers can point to exact files and line ranges.

Agentic RAG (Optional)

Enable agentic retrieval for complex, multi-hop queries or large repos.

Evaluation & Health

Quality scoring and health checks surface missing keys, slow sources, and low relevance.

Search tips

Use short keywords, then filter by source or tag when results are broad.

Repository Ingestion

ChipOS can ingest GitHub repositories as code-aware knowledge sources, storing snippets with file paths, line ranges, and language metadata.

Code-Aware Chunking

Defaults to ~200 lines per chunk with overlap for context.

Rich Metadata

Stores repo, path, language, and line range for citations.

Progress Tracking

Monitor ingestion progress and retry safely when needed.

Private repos

Add a GitHub token in Settings to crawl private repositories securely.

RAG Health & Evaluation

ChipOS evaluates RAG quality after ingestion and recrawls, surfacing issues like missing keys, low relevance, or slow sources before they impact users.

Healthy

75-100% score

RAG working well

Needs Attention

Warnings detected

Check providers or crawl status

Unhealthy

<50% score

Critical problems

Common signals

Low relevance scores, provider errors, or stalled crawls will lower health until fixed.

System Architecture

ChipOS Desktop App Desktop Shell
Workspace Selector
Dashboard Webview
Chat/Webviews
Main Server

Port 8181 • FastAPI + Socket.IO

MCP Server

Port 8051 • HTTP-based MCP

Agents Service

Port 8052 • Agent workflows

Chat (ChipChat)

Port 3080 • Fullscreen chat UI

Zendesk MCP Support automation
Tickets
Comments
Help Center
Managed Postgres PostgreSQL 15+ with pgvector
Crawled Pages
Code Examples
Embeddings
Relationships

Claude Code & MCP

ChipOS connects to external assistants through MCP. Use the Controls panel to copy configs for Claude Code, Cursor, and Windsurf, or to start Claude Code directly.

Claude Code

Auto-installed commands and agent packs

Fully Supported

Cursor

MCP server configuration for context injection

Fully Supported

Windsurf

Native MCP tool access for knowledge queries

Fully Supported

Other MCP Clients

Connect any MCP-compatible tool or IDE

Supported
Claude Vibe Mode

ChipOS can launch Claude Code in Vibe Mode for uninterrupted sessions when enabled.

Commands Reference

/axe <prompt>

One-shot pipeline builder with required output directories.

/axe Build an RTL lint pipeline for this repo
/init-chipos <workspace>

Initialize ChipOS tracking in the current workspace.

/init-chipos Setup this workspace for ChipOS
/voyager-help <query>

Ask the ChipOS knowledge base for contextual help.

/voyager-help How do I write a UVM sequence?
/build <target>

Build project artifacts or pipeline outputs.

/build rtl

API Reference

The ChipOS backend exposes REST and MCP APIs for integration.

REST Endpoints

Method Endpoint Description
GET /api/health Health check
POST /api/rag/query RAG query with context
GET /api/sources List indexed sources
POST /api/sources/crawl Start web crawl
POST /api/repositories/index Repository ingestion (code-aware)
GET /api/agent-work-orders List agent work orders

Settings Overview

ChipOS settings cover AI providers, RAG tuning, backend mode, and local runtime options.

Providers

  • Provider Status - OpenAI, Anthropic, Google, OpenRouter
  • Default Model - Per workflow or chat mode
  • Fallbacks - Configure routing preferences

API Keys

  • OpenAI - RAG queries & embeddings
  • Google - Alternative LLM provider
  • GitHub - Private repo crawling
  • Encrypted storage with show/hide toggle

RAG & Knowledge

  • Hybrid Search - Vector + keyword search
  • Reranking - Cross-encoder scoring
  • Agentic RAG - Advanced retrieval
  • Ingestion - Crawl and repo settings

Backend & Local

  • Hosted vs Local - Switch backend mode
  • Ollama - Validate local instances
  • Hosted Auth - Configure project and URL
Complete Settings Documentation

For detailed descriptions of all 50+ settings, recommended configurations, and troubleshooting tips.

Full Settings Guide

Troubleshooting

Login or Chat Mode Issues

If the login screen loops or Chat Mode is blank:

  1. Check your network connectivity
  2. Sign out and sign back in to reset your session
  3. Try restarting ChipOS with Cmd+Shift+P → Restart Services

MCP Server Not Connecting

If Claude/Cursor can't find the MCP server:

  1. Verify the backend is running (check Controls panel)
  2. Confirm the MCP URL (hosted vs local) in Settings
  3. Try restarting ChipOS with Cmd+Shift+P → Restart Services

Slash Commands Not Working

If /axe or /voyager-help doesn't respond:

  1. Check that .claude/commands/ exists in your workspace
  2. Re-open the workspace and re-select the agent pack
  3. Restart Claude Code or reload the IDE session

Empty RAG Results

If queries return no results:

  1. Check that sources are indexed in the Knowledge Base sidebar
  2. Verify embeddings are enabled and provider keys are set
  3. Try a broader query to confirm connectivity