Use CALIBER
Tools
The practical guide to defining, testing, governing, and using callable tools in workflows and assistant surfaces.
Use this page when you need the practical path for callable tools: define the tool, test it, wire it into a workflow or assistant surface, and understand the sandbox and approval boundaries.
At a glance
| Task | Start here | Deep reference |
|---|---|---|
| define or register a tool | shape the callable contract first | Tools architecture |
| verify tool behavior | use fixture-backed tests | Tools architecture |
| call the tool from a workflow | connect it through a workflow node | Workflows |
| govern side effects | inspect approval and runtime policy boundaries | Trust and governance |
1. What tools are for in CALIBER
Tools are versioned callable assets. They are not just snippets attached to a prompt. CALIBER tracks their contract, tests, and execution boundaries because they may touch real external systems.
2. Common tasks
| You want to... | Read this next |
|---|---|
| package a new callable | Tools architecture |
| use a tool in a flow | Workflows |
| expose the tool to an agentic surface | Aria assistant |
| audit why a tool call was blocked or paused | Trust and governance |
3. Common failure modes
| Symptom | First thing to check |
|---|---|
| The tool works locally but not in CALIBER | packaging, sandbox assumptions, or missing dependencies |
| The tool appears but cannot run live | approval policy or runtime permission boundary |
| The workflow compiles but the tool step fails | tool schema/inputs do not match the runtime call shape |