WorkAny

A desktop agent that turns natural-language requests into workspace, code, and tool operations.

Stars
★ 1.5k
Last updated
2mo ago
License
NOASSERTION
Primary language
TypeScript

At a glance

Works with
Universal · cross-platformOpenAI API · Claude API
You'll need
Node.js >= 20pnpm >= 9Rust >= 1.70Shell / CLINetwork accessLocal filesystemMCP Server
Typical use
An individual organizing files in a local workspace can submit the file-handling task in natural language.
Main limitation
Local development requires Node.js, pnpm, and Rust, creating more setup overhead than a browser-only tool.

What does this agent do, and when should you use it?

WorkAny is a desktop AI agent application that accepts natural-language tasks and streams execution in real time. Its repository describes a React 19, TypeScript, Vite, and Tailwind CSS 4 frontend; a Hono and @codeany/open-agent-sdk backend; and a Tauri 2, Rust, and SQLite desktop layer. The agent runtime runs in-process without an external CLI and exposes more than 30 built-in tools for file I/O, shell execution, web search, code editing, and related work. It can preview HTML, React, and code artifacts, while extending through custom skills and MCP servers over stdio, SSE, or HTTP. It fits users who want file work, code generation, and office-style output in one desktop workspace, provided they can support the required Node, pnpm, and Rust environment.

A user submits a natural-language task and WorkAny executes it with real-time streaming. Its Hono and @codeany/open-agent-sdk backend runs in-process and can use more than 30 built-in capabilities, including File I/O, shell execution, web search, and code editing; the desktop application manages the workspace and offers an isolated Sandbox for code execution. It can generate or handle code, HTML/React files, documents, data tables, and slides, with live Artifact Preview for HTML, React, and code files. Model access supports OpenRouter, Anthropic, OpenAI, and compatible API endpoints, while MCP servers can be integrated through stdio, SSE, or HTTP.

  1. An individual organizing files in a local workspace can submit the file-handling task in natural language.
  2. A developer prototyping a website can generate code and inspect HTML, React, or code artifacts in the application.
  3. An office user preparing a first draft of a document, data table, or slide deck can initiate the generation task from one desktop app.
  4. A team with existing MCP services can connect them to desktop task flows through stdio, SSE, or HTTP.
  5. A user working across OpenAI, Anthropic, OpenRouter, or a compatible endpoint can select a custom model provider for the agent.

How do you install or deploy this agent?

Requirements: Node.js >= 20, pnpm >= 9, and Rust >= 1.70. Install dependencies, then start the API and desktop application:

pnpm install
pnpm dev:api
pnpm dev:app

The README does not document where to enter model-provider credentials or the exact UI action for submitting a first task.

How do you use this agent?

Run pnpm dev:api, then pnpm dev:app to start the Web and desktop application. Enter a task in natural language; the README shows file organization, website generation, document generation, data-table generation, and slide generation as examples. For model access, use OpenRouter, Anthropic, OpenAI, or a compatible API endpoint. To extend the toolset, integrate an MCP server over stdio, SSE, or HTTP.

What are this agent's strengths and limitations?

Pros
  • The runtime uses @codeany/open-agent-sdk in-process, with no external CLI dependency stated in the README.
  • It consolidates more than 30 tools—including file I/O, shell execution, web search, and code editing—into a desktop task flow.
  • It supports OpenRouter, Anthropic, OpenAI, and compatible API endpoints rather than documenting only one provider.
  • It supports MCP servers over stdio, SSE, and HTTP, plus custom skills for extension.
  • It provides live Artifact Preview for HTML, React, and code files.
Limitations
  • Local development requires Node.js, pnpm, and Rust, creating more setup overhead than a browser-only tool.
  • The README does not specify API-key setup, provider configuration, or a default model.
  • The Sandbox is described as isolated, but its permission boundaries and security guarantees are not documented.
  • The README does not document failure recovery, task limits, resource limits, or concurrency behavior.
  • Licensing needs verification before adoption: repository metadata is NOASSERTION, while the README describes a WorkAny Community License with additional conditions based on Apache License 2.0.

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

Agent Source review Stars Updated Language Full support on
WorkAny This agent 18 · Major gaps ★ 1.5k 2mo ago TypeScript OpenAI API · Claude API
MyAgents 57 · Major gaps ★ 896 5d ago TypeScript Codex · Claude Code · Claude API
OpenHuman 51 · Major gaps ★ 40k today Rust
Skales 36 · Major gaps ★ 1.9k 6d ago ChatGPT · OpenAI API · Claude API

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Major gaps
18/ 100 5-point scale 0.9 / 5
Trust 0/29
Reliability 3/14
Adaptability 4/18
Convention 5/18
Effectiveness 4/13
Verifiability 2/8
Why each dimension lost points
Trust0 / 29 · 0.0/5

Evidence shows: The repository provides no documentation on permission model, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects control, rollback, or source attribution. README mentions sandbox and MCP support but lacks implementation details. Hence all trust criteria scored 0.

Reliability3 / 14 · 1.1/5

Evidence shows: README and package.json are consistent, but error handling documentation is missing. Dependency list is complete but availability guarantees are absent. Thus self_consistency and dependency_availability scored 1, failure_messages scored 0.

Adaptability4 / 18 · 1.1/5

Evidence shows: README describes multiple use cases (file organization, website generation, etc.) but does not define capability boundaries or trigger precision. Environment fit is supported by multi-platform build scripts. Hence audience_and_scenarios, capability_boundaries, environment_fit scored 1, trigger_precision scored 0.

Convention5 / 18 · 1.4/5

Evidence shows: README provides project structure, installation and development guides, consistent naming, example screenshots, but lacks known limitations and FAQ. License is a custom community license, version exists, CI/CD and contributing guide present. Thus most criteria scored 1, known_limitations scored 0.

Effectiveness4 / 13 · 1.5/5

Evidence shows: README displays output examples (screenshots) but no actual output quality or cost-benefit analysis. Hence output_usability, marginal_value, cost_benefit all scored 1.

Verifiability2 / 8 · 1.3/5

Evidence shows: Feature claims in README partially correspond to dependencies in package.json, but external verification is lacking. Hence claim_traceability and fact_inference_separation scored 1, cross_source_corroboration scored 0.

Risks and how to mitigate them
  • 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.
  • No permission model or user confirmation mechanism is provided, potentially allowing the agent to perform arbitrary actions.
  • Data flow and sensitive data handling are not documented, posing privacy risks.
  • Dependency security is not audited, and a custom license is used; careful evaluation is needed.
  • Known limitations and FAQ are missing, which may lead to misunderstandings about capability boundaries.
Evidence confidence: Low Reviewed Aug 09, 2026 Reviewed revision 7bc87487a18f
See the full review method →

FAQ

Must I use a particular model provider?
No. The README lists OpenRouter, Anthropic, OpenAI, and compatible API endpoints, but does not say where credentials are configured.
Does it depend on an external agent CLI?
No. The README says the @codeany/open-agent-sdk runtime runs entirely in-process with no external CLI dependency.
Can it affect local files or run system commands?
The README explicitly lists File I/O and shell execution, and describes an isolated Sandbox. It does not define the exact permission scope.
Can it connect to an existing MCP service?
Yes. The README states that MCP servers can be integrated through stdio, SSE, and HTTP.
How should I evaluate cost and licensing?
The README does not document model or service pricing. It calls the license the WorkAny Community License, based on Apache License 2.0 with additional conditions, while repository metadata is NOASSERTION; verify the LICENSE file before adoption.
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related agents