Overlay Workspace
Bring people, AI agents, shared knowledge, and connected tools into one workspace for delegated work and review.
- Source repo
- LayerNorm/overlay-web
- Stars
- ★ 197
- Last updated
- today
- License
- AGPL-3.0
- Primary language
- HTML
- FA score
- 59/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- High · needs real infrastructure
- You'll need
- Typical use
- A researcher who needs project materials, notes, and preferences kept together can save context and ask an agent to work from it.
- Not a fit if
- Teams seeking a ready-to-use workspace without setup
- Deployers unwilling to configure WorkOS, Stripe, or Convex
- Website
- getoverlay.io ↗ set on GitHub
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-20latest release v0.1.0-rc.1 (2026-07-16) · pre-release
- Source review
- 59/100 · Major gaps
What does this agent do, and when should you use it?
Overlay is an open-source workspace where people and AI agents share context, with Chat for conversation and Work for tool-driven execution. Users organize chats, notes, files, and memories in projects, then retrieve saved context to support later work. Agents can use browser sessions, cloud Computers, MCP servers, and external apps connected through Composio; users can review sources, files, and tool activity in the workspace. The interface is a Next.js 15 web app, backed by Convex functions, data, and realtime sync, with support for multiple model providers. The README links to a hosted product and documents a self-hosted development path; local setup requires Node.js 22, npm, a Convex account, WorkOS credentials, and Stripe test credentials.
Users upload files, create notes and memories, and organize them in projects; Knowledge supports semantic and keyword retrieval across files and saved context. They use Chat for multi-model conversations or Work to ask an agent to call tools, browse, or carry out a task. Available operations include interactive browser sessions, persistent cloud Computers with a terminal and desktop, MCP servers, and Composio connections to apps such as Gmail, Calendar, Notion, and GitHub. The workspace also supports image and video generation, audio recording and transcription, and scheduled automations that run at intervals or daily, weekly, or monthly. After execution, users can inspect sources, files, tool activity, and generated outputs, then continue in the same context.
- A researcher who needs project materials, notes, and preferences kept together can save context and ask an agent to work from it.
- A team with recurring procedures can schedule AI workflows to run at intervals or on a daily, weekly, or monthly cadence.
- A user who needs an agent to interact with web pages can run browser tasks in Work mode and inspect tool activity.
- A team connecting AI work to office apps can use Composio integrations for Gmail, calendars, Notion, or GitHub.
- A user reviewing agent output can inspect sources, files, tool activity, and generated images or videos in one workspace.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Config files
.env.example.env.local- Runtime
- Node.js 22
How do you install or deploy this agent?
Local development requires Node.js 22, npm, Convex development and production deployments, WorkOS authentication credentials, and Stripe test credentials. Copy the sample environment file, fill in the required configuration, and start the app:
npm install
cp .env.example .env.local
npm run devThe complete environment and deployment configuration is not included here; the README points to the project's local-run and self-hosting configuration docs.
How do you use this agent?
Start the web app and open the workspace, then upload files, save memories, or create notes and organize them by project. Choose Chat for conversation or Work to have an agent use tools, a browser, or connected apps; review sources, files, and tool activity after execution. Configure an integration before asking an agent to use an external app. To run recurring work, create an automation and select an interval, daily, weekly, or monthly schedule. The README does not provide a copyable API call or first-task command.
What are this agent's strengths and limitations?
- Chat and Work share a thread and context, so users can move from discussion to tool execution and review.
- Memories, files, notes, and projects live in one workspace with semantic and keyword retrieval.
- The repository documents a multi-provider model path alongside browser, cloud Computer, MCP, and Composio integrations.
- The workspace includes scheduled automation, image and video generation, and voice transcription.
- Local setup involves several prerequisites, including Node.js, npm, Convex, WorkOS, and Stripe.
- Browser, Computer, external-app, and media capabilities depend on corresponding runtime, integration, or provider configuration; availability varies by deployment.
- The README does not specify a fixed model catalog, model prices, or provider availability for each deployment.
- The project is under active development, and hosted and open-source versions may change; production, billing, and provider settings are environment-specific.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Overlay Workspace This agent | 59 · Major gaps | Web appFreemium | ★ 197 | today | HTML | — |
| Row-Bot | 76 · Good | Desktop appFree + model costs | ★ 1.7k | today | Python | ChatGPT · Codex · Claude.ai · OpenAI API · Claude API |
| Bitterbot Desktop | 55 · Major gaps | CLIFree + model costs | ★ 2.5k | 1d ago | TypeScript | OpenAI API · Claude API |
| Markus AI Workforce | 51 · Major gaps | Desktop appFree + model costs | ★ 219 | today | TypeScript | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
The protocol defines selected-root or all-user-files grants, scoped and expiring credentials, signed challenges, and request proofs. The authentication guide recommends OAuth/PKCE and describes personal tokens and access levels. The FAQ says agents act with the requesting person's access, users can inspect tool activity, account deletion requires typing DELETE, and archived items can be restored. This supports ordinary-use controls for permissions, confirmation, data-flow disclosure, and recovery. Deductions: the samples do not show the full enforcement path or confirmation behavior for every external action; data goes to the chosen model provider, whose retention rules vary; dependency-security evidence is limited; publisher identity and a verifiable source for it are unclear.
The protocol includes version fields, strict schemas, and request and batch size limits. The host workflow specifies cross-platform compatibility checks, and the README links to deployment and feature-status documentation. These provide some evidence of consistency. Deductions: the samples do not show how these checks cover the whole product or how dependency failures are handled in practice; availability of external services and required accounts depends on deployment; the included routes are forwarding layers and provide little evidence about the quality of failure messages.
The README describes human-agent collaboration and distinguishes Chat from Work. The protocol declares adapter, recovery, platform, filesystem-grant, and concurrency capabilities. The FAQ and authentication guide also explain personal versus shared spaces and access methods. Deductions: the samples do not fully define the boundaries of every adapter, feature flag, or runtime environment; scenario coverage is useful but not detailed enough to establish fit for every workflow.
The README clearly organizes documentation, features, repository layout, setup, and commands, with links to quickstart, API, configuration, and security material. The FAQ addresses common account and feature questions. The LICENSE text and package metadata both identify AGPL-3.0-only. Deductions: setup requires configuring services such as Convex, WorkOS, and Stripe, with instructions spread across linked documents; the sample cannot establish repository-wide naming or version stability; the project is described as actively developing, but the supplied material has no clear changelog or named maintenance responsibility and update path.
The product describes shared context, persistent memory, file retrieval, tool execution, integrations, and automation, and says users can inspect sources, files, tool activity, and outputs. This supports some practical output usability and collaboration value. Deductions: these are primarily product claims in documentation; the evidence does not quantify added value, operating cost, or setup burden, and the broad feature set depends on external accounts and services.
The README points to FEATURES.md for release status and to the deployed model selector for provider details; it says API documentation is generated from shared schemas. Protocol fields and workflow configuration provide implementation details that can be checked, while the FAQ makes explicit claims about visibility and provider data transfer. Deductions: this packet is only a sample, and some product claims lack corresponding implementation excerpts; several security and utility statements remain documentation claims without sufficient corroboration in the supplied material; facts and inferences are generally distinguishable, but evidence coverage is incomplete.
- Agents inherit the access of the person talking to them, and workspace content is visible to workspace members. Check connected apps, workspace scope, and granted access before use.
- Messages are sent to the selected model provider, and retention rules vary by provider. Self-hosted deployments also depend on external service configuration.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Overlay self-hosting release candidate This release candidate packages the provider-neutral Overlay runtime for an EC2 sanity-check deployment. Recommended deployment - Guide: https://www.getoverlay.io/docs/deploy-operate/docker-ec2 - Runtime image: ghcr.io/layernorm/overlay-web:0.1.0-rc.1 - Immutable digest: sha256:a21dd700fd2c84255e44b15dbcddb4e5b9d12502b325e4d4538f0e0b422d963c - Source commit: f96f3177b1aca4cca017e64e93e18481b2054b0d …View on GitHub ↗
FAQ
Is Overlay free to use?
What services and credentials are needed to self-host?
Which external tools can an agent use?
Do scheduled workflows and external actions run automatically?
Is Overlay Workspace open source?
Is Overlay Workspace safe to use?
Is Overlay Workspace still maintained?
Who makes Overlay Workspace?
What is Overlay Workspace's official website?
Thanks — we use this to improve the page.