Your tools, connected safely

Integrations in Runwork connect your workspace to external services like CRMs, payment processors, communication tools, and databases. In most teams, every AI tool on every computer holds its own copy of every connection: HubSpot OAuth in Claude Desktop, again in Cursor, again in ChatGPT, again on the next machine. Runwork keeps the credentials instead. Each person connects their own accounts once, sees what their teammates have connected, and their AI can then work in those tools from whichever AI tool they use. Inside a space, the apps, automations and AI agents reach the same services.

Adding an integration takes seconds. Find the service you need and complete the OAuth flow or enter your API credentials. Runwork securely stores these credentials and manages token refresh automatically. No more dealing with expired tokens or OAuth callback URLs in your application code. And it all happens without an engineering ticket: a marketing ops person can connect the whole stack themselves, and when a tool isn't in the catalog, their agent can still reach its API through a small workspace app it makes.

Connected services automatically become available as tools for your AI agents. If you've connected HubSpot, your agents can create contacts, update deals, or query your pipeline, just by asking. If you've connected Slack, agents can post messages or create channels. There's no additional configuration required. Underneath, this is MCP: every connected integration becomes a tool in your workspace's MCP server, which is how every AI tool your team uses gets to call them. Your AI agent can search the catalog, list connections, and call integration APIs on your behalf. From the terminal, runwork integrations call lets you proxy API requests through any connected service.

For automations and workflows, integrations provide trigger sources and action destinations. Receive a webhook from Stripe when a payment succeeds, then update your internal entities and send a Slack notification, all in one flow. The integration layer handles the API complexity while you focus on business logic.

Security is built in. You control which apps can access which integrations through scoping rules. Sensitive connections (like payment processors) can be restricted to specific apps. All API calls are proxied through Runwork with automatic rate limiting, retry handling, and error logging, none of which you have to build yourself. And because every call goes through one place, you can watch it: which integrations get used, by which apps and agents, and how often.

Integrations get useful when you combine them. HubSpot can sync to Slack the moment a deal moves, and a Stripe payment can trigger its own follow-up. Describe the combination to your agent and it makes the app. Runwork hosts it in the space, with the connections already in place.