> ## Documentation Index > Fetch the complete documentation index at: https://docs.engini.ai/llms.txt > Use this file to discover all available pages before exploring further. # Cookbook > Task-shaped recipes for connecting apps, running tools, building agents, and automating Engini. Recipes are organized by **what you're trying to do**, not by which surface you use - most show cURL, CLI, Python and TypeScript side by side, so pick whichever lane you're in. ## Connect apps Run the full OAuth handshake from inside your own product - the flow every embedded integration needs. Refresh the object catalog, poll until it settles, and choose which tables the connection exposes. ## Run tools Zero to executed tool with nothing but cURL. Walk every tool or application with offset/top, and survive rate limits. `isSuccess` vs HTTP status, reading `executionInfo`, and what each error code means. ## Build agents Decide exactly which tools an agent may call - enforced server-side, not by prompt. The canonical agent loop. **Start here** if you're new to agents. File uploads - the model picks attachments by name, your code controls the bytes. The same loop pointed at an ERP. ## React to events Destination, trigger, test event, and a receiver that verifies every signature. Read why attempts failed, bring a destination back, replay the gap, rotate a secret safely. Stream events to your laptop and rehearse signed deliveries on localhost - no tunnel. ## Automate The journey from the terminal - as a human, and as an agent using resumable gates. Headless auth, exit-code branching, and JSON output in a pipeline. Handles, dot-paths and projections - pay for the slice you need, not the payload. The SDK as a plain typed API client - the shape of a cron job. ## The two integration paths **High-level, with `Toolset`** - bind app → connection once, then two calls per turn: ```python theme={null} ts = client.toolset(connections={"salesforce": conn_id}) tools = ts.tools() # provider-shaped tool schemas results = ts.handle_tool_calls(reply) # resolves connection + runs each call ``` **Low-level** - drive the pieces yourself (fetch schemas, wrap, parse calls, execute, format results); see [LLM providers](/developers/sdk/llm-providers). Or skip the model entirely and call `client.tools.execute(...)` directly. Runnable sources for the agent recipes live in [engini-sdk/python/examples](https://github.com/engini/engini-sdk/tree/main/python/examples) - each is a self-contained script (`uv run `). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.