Workix

Workix

Languages: English · Русский · Español · Deutsch · Français · Türkçe · Українська · हिन्दी · 中文

Workix is a place where people, projects, and work find each other.

The Workix hub brings three things into one catalog:

• Projects — startups, products, and communities looking for people.
• Roles and orders — open work with a description, budget, tags, and a way to apply.
• Performers — specialists who are open to new projects.

Workix is not a traditional freelance marketplace. It does not try to replace the relationship between a project and a specialist. A listing leads to the project owner's preferred contact or application form, while Workix makes the opportunity easy to publish, discover, and share.

People can browse the catalog on the web. AI agents can search it and manage listings through MCP or the REST API. The same local MCP can also collect opportunities from external platforms while keeping platform credentials on the user's machine.

Open Workix: workix.co
Guide for agents: workix.co/agent
API reference: workix.co/api.txt

Ask your agent to explore Workix

Paste this prompt into Cursor, Claude, or another coding agent. It asks the agent to study Workix and then help with your specific goal.

Learn how Workix works:

1. Open https://workix.co/agent, https://workix.co/llms.txt,
   and https://workix.co/api.txt.
2. Open https://github.com/facetoplace/Workix and read README.md.
3. Explain in plain language:
   - what Workix is and who it is for;
   - how I can browse projects, roles, orders, and performers;
   - what an AI agent can do through the Workix MCP and API;
   - which parts run locally and how credentials are kept safe.
4. Look at my current project and help me choose and complete one useful next step:
   - connect Workix MCP to my agent;
   - search for relevant work or performers;
   - publish or update a project, role, or performer profile;
   - deploy the Workix storefront on my own domain;
   - contribute an improvement to Workix.

Use WORKIX_API=https://workix.co for the central hub.
Do not upload freelance-platform passwords or tokens to the hub.
Adapt the instructions to my operating system, tools, and project.

What is in this repository?

This repository contains the open clients and integrations for Workix. They use the central hub API at https://workix.co, so a local copy can display the shared catalog without maintaining a separate catalog.

• views/ contains the storefront pages.
• assets/ contains the browser application, styles, translations, PWA files, and public API documents.
• mcp/ contains the TypeScript MCP server, Workix tools, and adapters for supported job platforms.
• docker/ contains a small nginx image for serving the storefront.
• docs/ contains public guides for agents and self-hosting.

The two main parts are independent:

1. Storefront — a web interface for browsing the Workix catalog. You can use it at workix.co or host the UI on your own domain.
2. MCP server — tools that let an AI agent search Workix, manage your listings and profile, prepare proposals, and work with supported external sources.

Catalog data is served by the central Workix API. Production secrets are not required to run the storefront or connect the MCP client.

Use Workix with an AI agent

With WORKIX_API=https://workix.co, an agent can search projects, roles, orders, and performers. With a WORKIX_AGENT_KEY, it can also create or update listings and manage a performer profile.

Recommended: install MCP from this repository’s source (freshest tools and adapters). npm/npx is optional and may lag behind git.

Requires Node 22.5 or newer: the local store is SQLite through Node's built-in node:sqlite, so nothing has to compile at install time.

git clone https://github.com/facetoplace/Workix.git
cd Workix/mcp
npm install
npm run build

Example Cursor MCP configuration (from source):

{
  "mcpServers": {
    "workix": {
      "command": "node",
      "args": ["FULL/PATH/TO/Workix/mcp/dist/index.js"],
      "env": {
        "WORKIX_API": "https://workix.co",
        "WORKIX_AGENT_KEY": "wix_…"
      }
    }
  }
}

Optional shortcut: npx -y @workix/mcp · npm @workix/mcp · Official Registry co.workix/mcp · install UI: workix.co/agent

Register at workix.co to get an agent key. The key is optional for public search and required for actions tied to your account.

External platform credentials belong only in the local MCP environment. Do not send Upwork, Kwork, or other platform passwords and tokens to the Workix hub.

See mcp/README.md for installation, available tools, and source-specific settings.

Host the storefront on your domain

Projects and communities can serve the Workix UI from a domain such as work.example.com. The mirror still reads the shared catalog from https://workix.co.

1. Deploy views/ and assets/, or use the image in docker/.

2. Point the UI to the hub:

   <meta name="workix-api" content="https://workix.co" />
   

   You can also set API_BASE or WORKIX_API to https://workix.co.

3. Configure your DNS and hosting for the chosen domain.

Listings associated with that domain can be featured for visitors to the mirror. See work.facetoplace.app for a live example and docs/self-host.md for details.

Use other platforms through Workix MCP

Workix MCP is also a single interface for working with external freelance marketplaces and job boards. An agent can search several sources in one digest, open a specific listing, draft a proposal, and either submit it through an available API or prepare a browser checklist for you.

These integrations run locally. API tokens, platform logins, and browser sessions stay on your machine and are not sent to the Workix hub. Proposal submission always requires explicit human confirmation.

Platform readiness

The score describes how much of the intended Workix workflow is ready today, not how much of the platform itself is covered:

• 5/5 — search and submission work through a supported API.
• 4/5 — reliable search; submission or authentication still has a fallback or limitation.
• 3/5 — useful integration, but credentials, proxies, or an unofficial interface affect reliability.
• 2/5 — browser-assisted watch source rather than an automated feed.
• 1/5 — link/checklist only; deep automation is intentionally out of scope.

The Automation policy column is deliberately conservative. Platform terms and API permissions can change, so users must also follow the current rules of each source and their own account.

The current development order is:

1. Stabilize Freelancer.com and complete the Upwork OAuth/apply workflow.
2. Improve FL.ru, Freelance.ru, Weblancer, and Kwork reliability.
3. Add better filters for Remote OK and HH.ru, and grow the employer ATS company list.
4. Keep closed and ToS-sensitive platforms browser-assisted instead of building fragile scrapers.

About metered sources

Every source above is free to read except JobsPipe, which charges one credit per job it returns. Because of that it is the one board a plain include_jobs digest never touches — otherwise a scheduled digest could spend a monthly quota without anyone noticing. Reach it deliberately with the workix_jobspipe_search tool, by naming platforms: ["jobspipe"], or by opting in with JOBSPIPE_IN_DIGEST=1. A local counter tracks what this MCP spent and refuses to start a call once your configured monthly budget is gone; workix_jobspipe_usage shows what is left.

JobsPipe is read-only. It indexes other people's boards and has no endpoint for submitting a posting, so publishing a job through it is not possible — a listing reaches that index only by living on a source it already crawls.

About the JobSpy bridge

Indeed, Glassdoor, ZipRecruiter, Naukri and BDjobs are reached through JobSpy (MIT), which you install: pip install -U python-jobspy, Python 3.10–3.12. Nothing happens unless you name one of those platforms outright — a plain digest never touches them.

We call it rather than copying its code on purpose. Those boards are read through private endpoints using credentials taken from their own mobile apps; keeping that upstream means the credentials live in your installation instead of being redistributed inside this package, and the people who maintain those scrapers keep maintaining them.

Be aware that only Indeed returned results in our own testing. The rest are blocked or rate-limited from most addresses, and BDjobs currently raises an error inside JobSpy itself. Whether the others work for you depends on your network, and using them is your call under each board's terms.

Coverage today: 64 platforms in the catalog, 36 shipped as downloadable adapter modules. Beyond the job and freelance sources above, the same module system serves the dStore app catalog (publish a live site or PWA, find similar apps) and an optional Telegram channel reader.

The source catalog is mcp/platforms.json. Run workix_list_platforms for the machine-readable list and workix_sources_status to see which integrations are available with your current local configuration.

Documentation and feeds

• Agent guide
• Machine-readable overview
• API reference
• OpenAPI document
• Support
• RSS: orders, projects, performers

Contributing — help grow the job & channel coverage

Workix gets better when the community points us at more places to collect work.

Please open a GitHub Issue or PR if you know:

• a website / job board / hiring feed (public API, RSS, or a clear HTML list) worth wiring into MCP digests;
• a Telegram channel or chat where people regularly post IT/mobile/freelance/startup vacancies (public @username only).

Even a short note helps: URL or t.me/…, what niche it covers, and that posts are recent. You do not need to ship a full adapter — a proposal is enough. PRs that add an adapter or a channel to mcp/telegram-channels.example.json are especially welcome.

Also welcome: MCP tools, presets, tests, storefront improvements, docs, and translations.

Start with CONTRIBUTING.md · Telegram catalog: mcp/TELEGRAM-CHANNELS.md. Product questions: Workix support.

License

See LICENSE. Modification and redistribution require clear attribution. Commercial reuse requires prior agreement.