Skip to content
Aether Platform

Aether Workspace

Aether Workspace is the core development experience of Aether Platform, combining a full-featured Code-Server (web-based VS Code) with an intelligent Code Indexer (GrepAI) to deliver a truly AI-native IDE.

It’s not just “VS Code in the browser.” It’s an IDE that understands your code.

What is Aether Workspace?

Code-Server

Full VS Code in the browser — extensions, terminal, debugger, and Git. No installation, no local setup. Just open a URL and start coding.

Code Indexer (GrepAI)

Semantic code intelligence powered by GrepAI. Your codebase is automatically indexed, enabling AI agents to find relevant code on the first try — instead of burning tokens on blind grep and read loops. Up to 80% reduction in LLM token consumption.

Architecture

Aether Workspace runs inside each developer’s Workspace Pod. Code-Server serves the VS Code frontend, while the Code Indexer continuously maintains a semantic understanding of the codebase.

graph LR
    Dev[Developer]

    subgraph "Workspace Pod"
        CS[Code-Server\nWeb-based VS Code]
        Ext[AetherPlatform\nVS Code Extension]
        IDX[Code Indexer\nGrepAI Sidecar]
    end

    subgraph "AI Layer"
        Agent[AI Coding Agent\nClaude / Cursor / etc.]
        MCP[MCP Protocol]
    end

    Dev -->|Browser / Local VS Code| CS
    CS --- Ext
    Ext -->|REST API / MCP| IDX
    Agent -->|MCP Tools| IDX
    IDX -->|Semantic Search\nCall Graph| MCP

Component Roles

ComponentRoleTechnology
Code-ServerWeb-based VS Code with full extension supportcode-server
AetherPlatform ExtensionBridges VS Code UI with Code Indexer, provides commands and statusVS Code Extension API
Code Indexer (GrepAI)Semantic indexing, natural language search, call graph analysisGrepAI + FastAPI sidecar

Code Indexer: Cut AI Agent Costs by 80%

Without an index, AI agents blindly scan your codebase — reading file after file, running grep after grep — burning through tokens to find what they need. The Code Indexer eliminates this waste.

By pre-indexing your codebase semantically, agents locate the right code on the first query. The --compact output mode further compresses results, delivering only essential context. The result: up to 80% reduction in LLM token consumption per task.

Search your codebase by describing what the code does, not memorizing exact names.

Terminal window
# Find authentication logic — no need to remember function names
grepai search "user authentication flow" --compact


# Find error handling patterns across the project
grepai search "error handling middleware" --compact


# Locate database connection management
grepai search "database connection pool" --compact

Call Graph Tracing

Understand function relationships before making changes.

Terminal window
# Who calls this function?
grepai trace callers "HandleRequest"


# What does this function call?
grepai trace callees "ProcessOrder"


# Full dependency graph
grepai trace graph "ValidateToken" --depth 3

Auto-Indexing

The Code Indexer watches your workspace in real-time. Every file save triggers incremental re-indexing, keeping the search index perpetually fresh.

  • Indexed files & chunks: Tracked via /health endpoint
  • Watch daemon: Automatically runs in sidecar mode
  • Kubernetes-ready: /ready endpoint for readiness probes

Dual-Mode Operation

Aether Workspace adapts to your deployment context automatically.

Sidecar Mode (JupyterHub / Cloud)

In cloud deployments, the Code Indexer runs as a sidecar container alongside Code-Server. This is the primary mode for Aether Platform.

  • Auto-detected via GREPAI_INDEXER_URL environment variable
  • REST-based health monitoring (default: port 3200)
  • Integrated with Kubernetes orchestration

Local Mode (Development)

For local development or standalone use, the GrepAI MCP server spawns as a local process.

  • Communicates via MCP (Model Context Protocol) over stdio
  • Managed by the AetherPlatform VS Code Extension
  • Auto-start on workspace open

AI Agent Integration

Aether Workspace is designed to be the perfect host for AI coding agents. The Code Indexer exposes its intelligence via the MCP protocol, giving agents deep code understanding.

MCP Tools for Agents

ToolDescription
grepai_searchSemantic code search with natural language queries
grepai_trace_callersFind all callers of a function
grepai_trace_calleesFind all callees of a function
grepai_trace_graphBuild complete call graphs
grepai_index_statusCheck index health and statistics

The Cost Problem with AI Agents

Without Code Indexer, every AI agent session follows the same expensive pattern:

Agent: "Let me find the auth logic..."
  → grep "auth" (500 files scanned, 2000 tokens)
  → read file A (800 tokens) — not relevant
  → read file B (1200 tokens) — not relevant
  → read file C (600 tokens) — found it!
  Total: ~4,600 tokens just to locate one function

With Code Indexer:

Agent: grepai_search "authentication flow" --compact
  → 3 precise results with context (400 tokens)
  Total: ~400 tokens — 91% savings

At enterprise scale with dozens of agents running daily, this translates directly into significant LLM API cost reduction.

VS Code Extension Commands

The AetherPlatform Extension provides convenient commands for managing the Code Indexer:

CommandAction
Aether: Start GrepAIStart the Code Indexer
Aether: Stop GrepAIStop the Code Indexer
Aether: Rebuild IndexRebuild the workspace index
Aether: Index StatusShow indexing statistics

Access via the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) or the @aether chat participant.

Comparison: Traditional IDE vs Aether Workspace

CapabilityTraditional Cloud IDEAether Workspace
Code editingVS Code in browserVS Code in browser
Code searchText-based (grep)Semantic (natural language)
Code navigationSymbol-based (LSP)Symbol + Call graph tracing
AI contextRaw file accessPre-indexed semantic context
Agent token costHigh (blind scanning)Up to 80% reduction
Agent toolsManual setup per agentMCP protocol (universal)
Index freshnessManual reindexAuto-watch (real-time)
DeploymentSingle containerDual-mode (sidecar / local)