Tools & Actions
Tools are what make agents actually useful. A model without tools can only think and write. A model with tools can browse the internet, run code, send emails, update spreadsheets, call APIs, and interact with the real world.
Common tools
Web search & browsing
The agent can look things up, read articles, check prices, monitor news — anything you'd do in a browser.
Use case: Research an industry before a sales call. Monitor competitor pricing. Summarize recent news about a client.
Code execution
The agent writes and runs code in a sandboxed environment — Python, JavaScript, shell scripts. It can manipulate data, run calculations, and generate charts.
Use case: Analyze a CSV file, build a data model, automate a repetitive script.
File operations
Read PDFs, edit spreadsheets, process images, extract text from documents.
Use case: Process uploaded invoices, summarize a PDF report, update a shared document.
API calls
Connect to any external service — your CRM, accounting software, email platform, Slack, calendar.
Use case: Update a HubSpot contact, create a Jira ticket, send a Slack notification, book a calendar slot.
Memory read/write
Store and retrieve information from the agent's long-term memory.
Use case: Remember a customer's preferences, recall the context from last week's meeting.
Sub-agent spawning
Spin up another agent to handle a sub-task, then collect the result.
Use case: Manager agent delegates research to a specialist agent, compilation to a writing agent.
Tool safety
Tools are powerful — which means they can also cause real damage if misused or manipulated. A few principles:
- Least privilege — give the agent access to only the tools it needs for the task
- Human-in-the-loop for irreversible actions — sending emails, making purchases, deleting data should require confirmation
- Audit logs — always know what tools were called and with what parameters
This is exactly why the Steward concept exists — a local secrets vault that controls which tools an agent can access and prevents credential leakage even if the agent is manipulated.
Next: see these concepts in action → Recipes