Wave Engine — Your AI Engineering Partner in VS Code
Local-first AI that codes, deploys, monitors, and runs your day.
A multi-agent assistant that lives entirely on your machine. Chat with your codebase, ship it to production, watch it stay up, and get a morning briefing about everything else that matters — all without your data leaving the box.
Why Wave is different
Most AI extensions do one thing: complete code, or chat, or generate. Wave does all of it — plus everything the other tools ignore.
- Local-first brain. Bundled Qwen 2.5 Coder runs on CPU. No API key required. No per-token cost. Works offline.
- Reads your repo before answering. Every response is grounded in your real files with real line numbers — no hallucinated paths.
- Safety-first automation. Wave never invents facts. When it drafts a job application or fills a form, missing information stops the flow and asks you.
- Cross-session memory. Everything Wave learns about you and your projects sticks around. Your fixes, decisions, and preferences build up over time.
- Life OS. Wave doesn't just help you code — it aggregates your calendar, email, GitHub, and watchers into one morning briefing.
Command Center — one screen, every feature
A hub panel shows every Wave capability grouped by category (Chat, Code, Life OS, Personal AI, DevOps, Operations) with a live status strip:
- Brain — local model up/down
- Deploys — history + last-ok badge
- Uptime probes — failing / total
- API cost — this month, colour-coded against your budget cap
- Job applications — needs-info + ready-to-submit
- Briefing — cached vs. stale
- Scheduled jobs & watchers — live counts
Click any tile → jump straight to that dashboard. Every card is live-polling.
The full feature map
💬 Chat & code
- Multi-tab conversations — each chat opens in its own editor panel with auto-generated title
- AI Edit In-Place — Composer-style edits streamed into your editor
- Inline Edit (Ctrl+K) — Cursor-parity inline diff with accept/reject
- Debug Current File — Wave sets a breakpoint, launches the debugger, narrates each step
- Scaffold Wizard — interactive project generator (stack / auth / DB / deploy)
- Prompt Library — save + reuse parameterised prompts across projects
- Morning Briefing — Wave scans your email, calendar, GitHub notifications, watcher events, and recent job runs → hands them to local Qwen → concise 6-bullet digest
- Pre-Meeting Brief — auto-generated ~10 min before a meeting: WHO / LAST TIME / OPEN ITEMS / SUGGESTED TALKING POINTS
- Scheduler — cron + human-friendly schedules (
every 30m, 0 8 * * *). Visual builder with mode tabs (Every / Daily / Weekly / Custom). Live "next run in Xm" countdowns. One-click templates for common jobs. Auto-starts on boot.
- Watchers — background change detection for URL content, JSON paths, RSS feeds, file mtimes, keyword hits. Alerts via Slack/Discord/email.
- MCP Marketplace — curated catalog of 13 Model Context Protocol servers (Filesystem, GitHub, Postgres, Notion, Linear, Slack, Playwright, and more). One-click install with per-server required-arg forms — no hand-editing
mcp_servers.json.
- Second Brain — auto-indexed persistent memory. Every note, fix, decision, meeting, chat gets embedded via BM25. Search across everything: "what did I decide about retries last month?"
- Cross-Repo Semantic Search — register any number of repos on your disk. Search once, get hits from all of them. "which of my 30 projects had that rate-limiter fix?"
- Unified Inbox — aggregates email, GitHub notifications, calendar events, watcher events. Priority-ranked. Read / snooze / archive with persistent state.
- Cost & Budget Dashboard — per-query token & $ tracking across 14 model pricings. Monthly rollup by feature. Set a cap; queries auto-refuse near limit.
- Job Search + Auto-Apply — search RemoteOK / WeWorkRemotely / LinkedIn RSS. Wave assesses fit, drafts a cover letter + tailored resume, identifies form fields, stops at any missing info and asks you rather than guess. You approve before submit.
- Codebase Tour — point Wave at any repo, get a 3-page onboarding guide (architecture, top files in order, gotchas). Written to
<repo>/.masi/CODEBASE_TOUR.md.
- Env Snapshot & Diff — one click captures Python/Node/OS/packages/env vars/hardware into a JSON. Share with a colleague, load it, see a diff. Kills "works on my machine."
- Secret Scanner — 15+ regex patterns for AWS keys, GitHub PATs, Anthropic/OpenAI/Google keys, JWTs, PEM headers, Slack tokens, Stripe live keys. Catches leaks before commit.
- Session Handoff — encrypted (AES-256 / PBKDF2) bundle of every Wave state file. Drop into Dropbox/iCloud/USB, unlock on any other machine, continue mid-thought.
🚀 DevOps (deploy + monitor)
- Deploy to Vercel / Netlify / Render with detected provider
- Deployment history — every deploy logged with git SHA/branch/env/URL. One-click rollback for supported providers.
- CI/CD generator — auto-generate
.github/workflows/*.yml for Python-pytest, Node-test, Docker-build-to-GHCR, deploy-to-Vercel, deploy-to-Netlify. Stack-detection suggests templates.
- Docker — build, tag, push OCI images without leaving VS Code
- Envs + Encrypted Secrets — dev/staging/prod variable sets. Secrets encrypted with Fernet + PBKDF2. Push to Vercel/Netlify CLIs from Wave.
- Uptime probes — HTTP / SSL cert expiry / DNS resolution. Cadence per probe. Live sparklines, uptime % (24h/7d), median + p95 latency, up/down streaks. Auto-runs via scheduler.
- Post-deploy smoke tests — health probes with auto-rollback on failure
- Log streaming — tails Vercel / Netlify / Render / Docker logs in the UI
- Alert dispatch — probe failures fire to Slack/Discord/email with 60-min dedup
📄 Documents workbench
- Create Excel / Word / PowerPoint / PDF files from a form
- CSV → Excel with header-bold + auto-sized columns
- Markdown-lite → Word (headings, bullets, paragraphs)
- Slide blocks (title + bullets, separated by
---) → PowerPoint
- Recent-files browser filtered per format
🔔 Notifications
- Persistent notification center with a status-bar bell (
◇ bell Wave: N)
- Aggregates probe failures, job apps needing your input, scheduler errors, deploy failures
- Toasts vanish; this doesn't — click a row to jump to the relevant dashboard
- Polls every 30s
📚 Artefact History
- One screen listing every env snapshot, codebase tour, session bundle, and secret-scan Wave has produced
- Diff any two env snapshots to see version + package deltas
- Open any file in one click
⚙️ Settings Hub
- Cost cap + alert % + hard-block toggle
- Auto-apply profile short form
- Jump to VS Code settings filtered by
masi
- No more hand-editing
~/.masi/*.json
🩺 Operations
- Health Dashboard — every subsystem's status with one-click fix links
- Agent Activity — live view of every agent, tool call, reasoning step
- Doctor — 30+ health probes across model / API / DB / network / credentials
- Feature Tour + Onboarding — 6-step first-run wizard plus a 20+-slide carousel
Supported LLM providers
Bring your own API key, or run local:
| Provider |
Models |
| Local (Qwen) |
Qwen 2.5 Coder 3B (bundled, ~1.9 GB, CPU-only) |
| Anthropic |
Claude 4.7 Opus, 4.6 Sonnet, 4.5 Haiku |
| OpenAI |
GPT-4o, GPT-4 Turbo, GPT-3.5 |
| Google |
Gemini 1.5 / 2.0 Flash + Pro |
| OpenRouter |
200+ models via single key |
| DeepSeek |
DeepSeek-Coder, DeepSeek-V3 |
| Together / Groq |
Meta Llama, Mistral, and more |
| OpenAI-compatible |
Any endpoint (/v1/chat/completions) |
Keys live in VS Code's encrypted secret storage. Local Qwen requires zero credentials. Cost dashboard tracks every paid query so you can see exactly what a session cost.
Getting started
- Install from the Marketplace
- Click the wave icon in the activity bar
- Follow the 6-step onboarding wizard — connects an API key or downloads the offline model
- Take the Feature Tour (or explore via Command Center — Ctrl+Shift+P → "Wave Engine: Open Command Center")
- Ask anything in plain English, or click any tile to open a dashboard
Command palette
Every feature is reachable from Ctrl+Shift+P:
Wave Engine: Open Command Center (Home)
Wave Engine: Meet Wave — Feature Tour
Wave Engine: Open Life OS
Wave Engine: Open Personal AI (Brain / Repos / Inbox / Cost / Jobs)
Wave Engine: Open DevOps (Deploy / CI/CD / Envs / Monitor / Logs)
Wave Engine: Open Documents Workbench
Wave Engine: Show Notifications
Wave Engine: Show Artefact History
Wave Engine: Open Settings Hub
Wave Engine: Show Health Dashboard
Wave Engine: Show Agent Activity
Wave Engine: Show Prompt Library
Wave Engine: AI Edit In-Place (Composer)
Wave Engine: Debug Current File with Breakpoint
Wave Engine: Deploy to Vercel/Netlify/Render
Wave Engine: Generate Codebase Tour for current folder
Wave Engine: Capture Environment Snapshot
Wave Engine: Scan current file for secrets
Wave Engine: Export encrypted session bundle
Wave Engine: Import session bundle
- …and 30+ more
Or use the status bar quick pick — bottom-right ◇ Wave item shows live counts and opens a searchable menu of every feature.
Privacy & data handling
- Local backend runs on
http://127.0.0.1:8001 — no telemetry, no analytics
- Local Qwen path runs entirely on CPU, no external calls
- Cloud model calls go directly to the provider you configure — Wave never proxies them
- Secret scanner + env-snapshot redaction ensure secrets never leave your machine
- Session bundles are AES-encrypted (Fernet + PBKDF2 200K iterations) with a password only you know
- Your code never leaves your computer unless you explicitly send it to a cloud model
Configuration
{
"masi.apiBaseUrl": "http://127.0.0.1:8001",
"masi.pythonPath": "/usr/bin/python3",
"masi.autoApply": false,
"masi.autoRunSafeCommands": false,
"masi.permissionMode": "balanced"
}
Or open the Settings Hub webview for a friendly form (Ctrl+Shift+P → "Wave Engine: Open Settings Hub").
Supported languages
Python · TypeScript · JavaScript · Go · Rust · Java · C / C++ · PHP · Ruby · Kotlin · Swift · Scala · Shell · SQL · Markdown · YAML · TOML · JSON
Cross-repo search indexes all of them.
System requirements
- VS Code: 1.96.0 or later
- Python: 3.12+ (for the backend)
- RAM: 4 GB minimum (8 GB recommended if running the local Qwen model)
- Disk: ~2 GB for the offline model
- OS: Windows 10/11, macOS (Apple Silicon + Intel), Linux (x64 / ARM64)
Optional (feature-gated):
docker CLI — for the DevOps Docker tools
vercel / netlify / render CLIs — for their respective deploys + rollback + log tail
cryptography Python package — for AES-encrypted secrets (falls back to base64 with a clear warning)
openpyxl / python-docx / python-pptx / reportlab — for the Documents workbench (per-format optional)
playwright — for browser automation via MCP or the web action tools
Wave never fails silently — every missing dependency shows up in the Health Dashboard with a one-click install command.
What's new in 4.6.0
- DevOps Dashboard — deploy history + rollback, GitHub Actions generator, encrypted secrets, uptime probes, log streaming
- Personal AI — Second Brain, cross-repo search, unified inbox, cost dashboard, auto-apply job workflow
- Documents Workbench — Excel / Word / PowerPoint / PDF UI (was action-only before)
- Notification Center with status-bar bell — persistent inbox for anything needing your attention
- Artefact History browser — visual list of every env snapshot / codebase tour / session bundle / secret scan
- Settings Hub — one form for cost cap, auto-apply profile, extension config
- Command Center upgrade — 8-card live status strip covering every subsystem
- Life OS extensions — visual cron builder, one-click job templates, live "next run" countdowns, auto-starting scheduler
- 7 new backend packages totaling ~2 000 lines of Python, all with API endpoints under
/api/v1/*
Full changelog → CHANGELOG.md.
Built by Aditya Talekar at Fundamental Labs.
| |