CrewAI lets you split complex workflows across specialized agents, but each agent still needs authenticated access to user tools like Gmail. This cookbook shows how to wire Scalekit OAuth into a CrewAI crew via MCP, then build a three-agent pipeline that scans, classifies, and drafts replies to a user’s unread emails.
Scalekit AgentKit returns native LangChain StructuredTool objects. Enable LangSmith tracing and every tool call — Gmail fetches, Slack messages, GitHub searches — appears as a traced span automatically. This recipe walks through setup, a working agent, and verifying traces in the LangSmith dashboard.
Poll Gmail with AgentKit tools, send each new thread through a multi-stage LiteLLM pipeline with per-stage models, notify Slack, then approve filing a GitHub issue and sending a reply from a localhost dashboard.
Apify Actors run in isolated containers with no persistent session — there is no concept of “who is logged in.” This recipe shows how to use Apify’s built-in user identity as the key into Scalekit’s OAuth token vault, so each user who runs your Actor connects their own third-party accounts, and tokens survive across runs without any user-managed input fields.
Machine-to-machine access tokens are JWTs signed by Scalekit. Your API verifies signatures using the published JWKS at /keys, then authorizes requests using the scopes embedded in each token. This cookbook ties those ideas together end to end.
