Runwork
For Engineers

Keep using your tools.
Share what you build.

You already use Claude Code or Cursor for everything. Runwork gives your team a shared workspace where your internal tools, agents, workflows, and data all live together. Build with your preferred AI. Your ops, marketing, and sales teams use what you create.

Get started Explore the CLI

One command connects 8 AI agents to your workspace

Claude Code Claude Code Claude Desktop Claude Desktop Cursor Cursor Codex Codex Windsurf Windsurf Copilot (VSCode) Copilot (VSCode) Copilot CLI Copilot CLI Gemini CLI Gemini CLI

You're already technical. You write real code with AI agents every day. The question isn't whether you can build things. It's where those things go, and who gets to use them.

With Runwork, you keep using Claude Code, Cursor, Codex, whatever you prefer. The platform keeps your local environment in sync with your workspace: knowledge files, framework docs, skills, MCP connections, app context. Your agent understands the full Runwork platform and your workspace from the first prompt. runwork dev maintains all of this dynamically as you work, and runwork sync keeps everything up to date across sessions.

What changes: everything you build lives in a shared workspace. That lead tracker isn't a local project. It's a running app your sales team can see on their dashboard, with its own MCP server any AI tool can connect to, a skill the workspace AI learns from, scheduled jobs, connected integrations, and public endpoints other systems can call. You write the code. Your whole team uses the result.

Skills and MCP servers sync across your team. When you push a skill or add an MCP server, every teammate's AI agent gets updated. Devs get the same context in their Claude Code. Non-devs get it in their Claude Cowork Desktop. Same workspace, same knowledge, different surfaces.

How it works

1

Install and connect

Install the CLI. Run runwork setup. It auto-detects your local AI agents and configures them with your workspace skills and MCP servers.

$ curl -fsSL https://runwork.ai/install.sh | sh
$ runwork setup
Detected agents:
✓ Claude Code
✓ Cursor
✓ Codex
✓ 12 skills synced
✓ 4 MCP servers configured
2

Build with your preferred agent

The platform keeps your local environment in sync with your workspace: knowledge files, skills, MCP connections, framework docs, and app context are all maintained dynamically during runwork dev. Your agent understands the full picture from the first prompt.

$ runwork init "Lead Tracker"
$ runwork dev
✓ Sandbox ready
✓ Knowledge files synced
✓ 12 skills available
✓ Preview → https://preview-a3x.runwork.ai
3

Deploy and your team sees it

One command to production. Your app shows up on the team dashboard. It automatically gets its own MCP server (so any AI tool can access it), auto-generates a skill (so the workspace AI learns it), and exposes public endpoints if you defined any. Frontend, backend, or both.

$ runwork deploy
✓ Deployed to production
✓ Dashboard: app visible to all teammates
✓ MCP server: live at /apps/lead-tracker/mcp
✓ Skill: auto-generated from app capabilities
✓ Endpoints: GET /api/leads, POST /api/leads
→ https://your-workspace.runwork.ai
4

Sync your whole team

Skills, MCP servers, and agent settings sync to every teammate. Devs get updated configs in Claude Code. Non-devs get the same context in Claude Cowork Desktop. Run runwork sync or let the desktop app handle it. Even outside Runwork codebases, your agent has full workspace context.

$ runwork sync
✓ 3 new skills synced (incl. Lead Tracker)
✓ 1 MCP server updated
✓ Claude Code config refreshed
✓ Cursor config refreshed
✓ 2 teammates notified

What your team gets from your work

Build frontends, backends, full-stack apps, standalone APIs, or any combination. Everything you deploy becomes part of your team's shared workspace.

Dashboard presence

Every app you deploy shows up on the team dashboard. Non-technical teammates can use it without knowing how it was built.

Auto-generated MCP server

Every app automatically gets its own MCP server. Claude Desktop, Cursor, and any MCP tool can read, write, and interact with your app's data and capabilities.

Auto-generated skill

Every app auto-generates a skill. The workspace AI and your teammates' agents learn what your app can do and how to use it.

Public endpoints and APIs

Define REST endpoints in your app. They're immediately live and callable by external systems, webhooks, or other apps in the workspace.

Agents in sandboxes

Your agents get their own computer. Delegate complex tasks to Claude CLI running in isolated environments with full compute.

Workflows and schedules

Recurring jobs, multi-step automations, cron schedules. All managed through the dashboard or CLI.

Shared data and files

Entities, file storage, and structured data your whole team can access. One source of truth across all apps.

3,200+ integrations

Connect HubSpot, Stripe, Slack, Google, and thousands more. Your agents and apps call them natively.

Frontend, backend, or both

Build a dashboard with a UI. Or a pure backend with API endpoints and scheduled jobs. Or a full-stack app with everything. Runwork runs all of it.

Your code. Your team's cloud.

Install the CLI and connect your AI agents in under a minute.

Then runwork login and runwork setup.