Overlay
An open-source control plane workspace where humans and AI agents share context, with persistent cloud desktops, multi-platform deployment, and governance controls.
At a Glance
Auto model and core workspace, no card required
Engagement
Available On
Alternatives
Listed Oct 2026
About Overlay
Overlay is an open-source workspace built by LayerNorm that lets humans and AI agents share context and get work done together. The core platform is licensed under AGPL-3.0 and the source code is publicly available on GitHub. It combines agent management, persistent cloud computers, multi-platform deployment, and workspace governance into a single interface.
What It Is
Overlay positions itself as a "control plane for your workspace" — a unified environment where you create, deploy, and manage AI agents alongside the ones you already use (such as Codex, Claude Code, Cursor, and Devin). Rather than treating AI as a chat box bolted onto existing tools, Overlay is designed around the full lifecycle of agent work: memory, tools, persistent computers, scheduling, and reach across platforms like Slack, Telegram, iMessage, and the web. The project describes itself as "purpose-built for a world where software works for you."
Architecture and Open-Source Model
The repository (LayerNorm/overlay-web) is a Next.js 15 application with a Convex backend for real-time sync, WorkOS for enterprise authentication, Stripe for billing, and Composio for 100+ external app integrations. Key technology choices include:
- Vercel AI SDK — AI streaming, tool calling, and multi-provider support
- Convex — Backend functions, data, and realtime sync
- Browser Use — Browser automation for interactive web tasks
- Daytona / Vercel Sandbox — Isolated code execution environments
- MCP servers — Connect external tool servers via Model Context Protocol
The web/server surface and a separate macOS Desktop client (Electron, Apple Silicon) are both publicly sourced. The license is AGPL-3.0-only, with paid commercial licenses available separately from LayerNorm Inc.
Agent Roster and Platform Reach
Every agent — whether created natively in Overlay or brought in via the Agent Client Protocol — appears in a single directory with its status, reach, and spend. Supported external agents listed on the homepage include OpenAI Codex, Anthropic Claude Code, Cursor, Devin, GitHub Copilot, Gemini CLI, and Ollama. Agents can be deployed to Slack, Telegram, iMessage, and the web from the same workspace without duplicating configuration.
Persistent Computers and Automation
A notable feature is persistent cloud desktops assigned to each agent — Ubuntu environments with signed-in apps, real browsers, and files that persist across sessions. Users can watch the agent's screen live. Alongside this, Overlay supports scheduled automations (interval, daily, weekly, monthly), spend caps per run, and durable workflow runs that report back on completion.
Governance and Workspace Controls
For organizations, Overlay surfaces spend limits, per-agent budget caps, tool approval requirements, model routing policy, and daily audit log exports. The manifesto published by LayerNorm emphasizes privacy, openness, control, and simplicity as architectural principles — including the ability to self-host the entire platform. A private/on-premises deployment path is available for organizations via a separate contact flow.
Current Status
The GitHub repository shows active development, with the last push recorded in late September 2026. The project notes that "the hosted product and the open-source repository may change as the collaboration model, provider integrations, and self-hosting support mature." A macOS Desktop client is available for download (Apple Silicon), described as being behind a "Gate B" release control. The repository had 166 stars and 22 forks at the time of indexing.
Community Discussions
Be the first to start a conversation about Overlay
Share your experience with Overlay, ask questions, or help others learn from your insights.
Pricing
Free
Auto model and core workspace, no card required
- Unlimited Auto model messages
- Notes, chats, files, knowledge
- Basic AI tools and core flows
- 10 MB file storage
Paid
For light monthly use with premium models and features
- Everything in Free
- Premium chat models
- Daytona sandboxes
- Browser tasks
- Image and video generation
- 1 GB file storage
Enterprise
Private/on-premises deployment with governed automations and custom storage
- Everything in Paid
- On-premises / private deployment
- Governed automations
- Custom file storage
- Curated model routing
Capabilities
Key Features
- Unified agent roster for native and third-party agents
- Persistent cloud desktops (Ubuntu) for each agent
- Live screen viewing of agent activity
- Multi-platform deployment: Slack, Telegram, iMessage, Web
- Scheduled and triggered automations
- Provider-neutral model support (OpenAI, Anthropic, Google, DeepSeek, xAI, Mistral, Kimi, Qwen, OpenRouter)
- Agent Client Protocol (ACP) compatibility
- MCP server integration
- Browser automation via Browser Use
- Code execution sandboxes (Vercel Sandbox, Daytona)
- Persistent memory and knowledge base with semantic search
- Rich notes editor with markdown and slash commands
- Image and video generation
- Voice input and transcription
- 100+ external app integrations via Composio
- Workspace spend limits and per-agent budget caps
- Tool approval controls and audit logs
- Self-hosting support
- Enterprise SSO via WorkOS
- Project organization for chats, notes, and files
