Everywhere Desktop Context Assistant
A desktop assistant that uses the active screen and app context to invoke multiple models and tools in place.
- Source repo
- Sylinko/Everywhere
- Stars
- ★ 6.3k
- Last updated
- 8d ago
- License
- NOASSERTION
- Primary language
- C#
- FA score
- 30/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude API
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A Windows user facing an error dialog can invoke the assistant next to it and ask what the error means and how to resolve it.
- Not a fit if
- Linux desktop users: the Linux build is still in development
- Teams needing fully offline operation: it relies on Everywhere Cloud or third-party model APIs
- Commercial teams wanting to fork and modify freely: it uses the Business Source License 1.1, not an open-source licence
- Source review
- 30/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
Everywhere is an interactive desktop AI assistant for Windows and macOS, built with Avalonia and marked as using .NET 10. It says it gathers structured data from the active software environment through screen multimodality, accessibility APIs, and UI automation. A user invokes it beside the current application with a global hotkey, then asks a question or works from selected text and context-aware strategies. Its documented tool set includes web browsing, sub-agent dispatch, local filesystem access, terminal script execution, MCP tools, Windows Everything Fast Search, and macOS system-app integration. The listed model options are Everywhere Cloud Service, OpenAI, Anthropic, Gemini, DeepSeek, Kimi, MiniMax, local Ollama, and custom API endpoints; Linux is still in development.
A user invokes Everywhere beside the current app with a global hotkey. It reads screen information and active-environment data obtained through accessibility APIs and UI automation, then accepts a typed request or selected text. It can produce an answer using a selected model provider or Everywhere Cloud Service and perform the documented operations Web Browser, Local File System, Terminal Script Execution, MCP Tools, and Dispatch Sub-agents; Windows additionally lists Everything Fast Search, while macOS lists integrated system apps. Results are presented in the desktop assistant as answers, summaries, translations, rewriting suggestions, or context-specific execution strategies. The supplied material does not document concrete API names, class names, CLI entry points, or tool-call arguments.
- A Windows user facing an error dialog can invoke the assistant next to it and ask what the error means and how to resolve it.
- A researcher reading a long technical web article can invoke it over the page and request a concise summary.
- A Chinese-speaking user working with foreign-language material can select text or invoke the assistant beside it to request a translation.
- A professional drafting a business email can ask for a more professional version while working in the draft window.
- A user assessing an uncertain online claim can select the relevant text and ask “Is it true?” for search and verification.
- An advanced user who has configured MCP tools or local models can combine desktop context with tools, filesystem access, and terminal scripts.
How do you install or deploy this agent?
The README provides release-package names and a download recommendation, but no source-build or command-line installation commands. Download the matching package from the official download page: Windows offers Everywhere-Windows-x64-Setup-vx.x.x.exe or Everywhere-Windows-x64-vx.x.x.zip; macOS offers Everywhere-macOS-arm64-vx.x.x.pkg for Apple Silicon and Everywhere-macOS-x64-vx.x.x.pkg for Intel Macs. Minimum requirements are Windows 10 10.0.19041.0 or macOS Monterey 12.0; Linux is unfinished. The README does not document provider credentials, MCP setup, permission-grant flow, or exact first-run configuration.
How do you use this agent?
After installation, invoke Everywhere beside the active application with its global system hotkey; the README does not state the actual key combination. Enter a request—for example, explain an on-screen error, summarize a page, translate text, or improve an email—or select text before asking. The README confirms support for particular models, Ollama, filesystem access, terminal scripts, MCP, and sub-agents, but supplies no copyable setup or invocation procedure for those features.
What are this agent's strengths and limitations?
- It explicitly combines screen context with accessibility APIs and UI automation, instead of relying solely on manual screenshots or copy-and-paste.
- Its listed model ecosystem spans OpenAI, Anthropic, Gemini, DeepSeek, Kimi, MiniMax, Ollama, its cloud service, and custom API endpoints.
- Its documented tool surface combines browsing, MCP, local filesystem access, terminal script execution, and sub-agent dispatch, with Windows and macOS-specific integrations.
- Global hotkeys, text-selection interaction, Markdown, and math rendering are designed for use within an existing desktop workflow.
- Linux support remains in development, so it is not yet a documented fit for a Linux desktop workflow.
- The README does not expose provider-credential setup, MCP server setup, permission boundaries, or safeguards around terminal execution; these need validation before deployment.
- Everything Fast Search on Windows and integrated system apps on macOS are platform-specific, so behavior may differ across operating systems.
- The supplied repository metadata says NOASSERTION while the README claims Business Source License 1.1; commercial adopters need to verify the actual license file and terms.
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 |
|---|---|---|---|---|---|---|
| Everywhere Desktop Context Assistant This agent | 30 · Major gaps | Desktop appFreemium | ★ 6.3k | 8d ago | C# | OpenAI API · Claude API |
| DeepChat | 66 · Some gaps | Desktop appFree + model costs | ★ 6.4k | 1d ago | TypeScript | OpenAI API · Claude API |
| PyGPT Desktop Assistant | 49 · Major gaps | Desktop appFree + model costs | ★ 2k | 1mo ago | Python | OpenAI API · Claude API |
| holaOS | 52 · Major gaps | Desktop appFreemium | ★ 11k | 9d ago | TypeScript | Codex · Claude Code · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: README claims support for multiple LLMs, MCP tools, local file system, terminal script execution, etc., but provides no details on permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. Deductions: All trust criteria lack evidence, and high-risk features (e.g., terminal execution) are mentioned without security considerations.
Evidence: README describes features but lacks architectural details or error handling information. Deductions: Self-consistency is based only on README description without code verification; dependency availability only mentions third-party services without dependency management details; failure messages are undocumented.
Evidence: README lists multiple use cases (e.g., troubleshooting, web summarization, translation, email drafting) and platform support (Windows, macOS), and specifies system requirements. Deductions: Capability boundaries are not clearly defined; trigger precision only mentions hotkeys and text selection; environment fit covers only Windows and macOS, with Linux in development.
Evidence: README is well-structured, including installation instructions, use cases, technical features, system requirements, contribution guidelines, and license. Deductions: Naming stability is not mentioned; known limitations are only partially mentioned (e.g., Linux in development); versioning/changelog is not provided; maintenance responsibility is only implied through contribution guidelines.
Evidence: README describes output as Markdown and math rendering, and use cases are clear. Deductions: Output usability lacks actual examples or user feedback; marginal value is based on claimed features; cost-benefit lacks pricing or resource consumption information.
Evidence: Claims in README (e.g., support for multiple models, platforms) are traceable to the document itself, but no independent verification is provided. Deductions: Claim traceability is based only on README; cross-source corroboration is insufficient; facts and inferences are not clearly separated.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- The agent has high-risk capabilities such as terminal script execution and local file system access, but no permission control or user confirmation mechanisms are provided, posing security risks.
- No data flow transparency is provided; users cannot understand how data is collected, used, and stored.
- Dependency security is not assessed; third-party dependencies may have vulnerabilities.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.9.0-canary.20261012.26 Pre-release
Release notes excerpt
v0.9.0-canary.20261012.26 — 2026-10-12 <img height="500" alt="banner" src="https://github.com/user-attachments/assets/0af1192d-4f44-47a6-a040-bb6d654c1d5e" /> _Created by @Requiem-OvO_ --- This small Canary update focuses on context compression, interface polish, and settings performance. 🚀 Improvements - Context Compression: Refined compression boundaries when sending, regenerating, and continuing responses, keeping the current user message intact until appropriate and ensuring tool calls and their results are handled together. …View on GitHub ↗ -
v0.8.2 Everywhere v0.8.2
Release notes excerpt
v0.8.2 - 2026-09-16 v0.8.2 makes long conversations easier to navigate and everyday chat workflows more flexible. This release introduces visual conversation turn navigation and first-class Mistral AI support, while improving search, attachments, file safety, web browsing reliability, and overall chat stability. ✨ Features & 🚀 Improvements …View on GitHub ↗ -
v0.8.1 Everywhere v0.8.1