Use cases
Each of these takes a job someone actually has - answering support questions, pulling data out of uploaded documents, researching a company - and builds it into something you could put in front of a colleague, starting from an empty config directory.
Real projects tend to need four or five capabilities at once, and the order you add them in changes how much trouble you have. Index and check a document set before anyone writes a prompt.
| Walkthrough | What you build | Capabilities it combines |
|---|---|---|
| Support agent | An agent that answers from your product documentation in Slack, and escalates rather than guessing | RAG, skills, memory, channels, autonomy |
| Document pipeline | A durable workflow that pulls structured data out of uploaded files and routes exceptions to a person | Attachments, structured output, workflows, human approval, artifacts |
| Research agent | Parallel investigation that returns a sourced brief and files it as a downloadable artifact | Subagents, web tools, parallel workflow nodes, artifacts, budgets |
How to read these
Every step says what it adds, so you can stop once you have enough. The first half of the support agent is a working agent on its own.
They assume you have an AXL server you can restart and a CONFIG_DIR you can edit. If you have
not created an agent before, Create your first agent makes all
three of these easier to follow.
Agent keys, tool names, and document sets in the examples are placeholders. Swap them for whatever exists in your deployment.
Next
Answers from your product docs in Slack, and escalates when it should not guess.
Document pipelineStructured extraction from uploads, with a person on the doubtful cases.
Research agentParallel investigation that returns a sourced brief.
ExamplesWorking configurations to paste in and adjust.