Guide
How agents should reason about peer-cli before they touch any command.
This section explains how peer-cli behaves as an operational surface.
It is the minimum context an agent should load before issuing write commands or wiring the CLI into a broader workflow.
Core Rules
- Prefer JSON output unless a human explicitly asks for table formatting.
- Assume every write is preview-first unless
--yesis set. - Prefer
peer mcpin read-only mode unless there is a specific need for write tools. - Treat configuration resolution as layered, not implicit.
- Use the catalogs for machine reasoning and the docs for workflow context.
Command Shape
Every command follows the same high-level form:
peer [global flags] <group> <subcommand> [typed flags]Examples:
peer quote --from USD --amount 100
peer deposit list --owner 0x1111111111111111111111111111111111111111
peer --env staging vault list --manager 0x1111111111111111111111111111111111111111Global Operator Defaults
The following defaults matter operationally:
| Setting | Default | Notes |
|---|---|---|
env | production | Can be overridden by CLI flag, env var, or stored config. |
format | json | Strongly preferred for automation. |
yes | false | Keeps write flows in preview mode. |
debug | false | Extra stderr logging for investigation. |
chain | Base / 8453 | Used throughout SDK-backed operations. |
Read vs Write Discipline
The command surface spans both pure reads and state-changing operations.
| Mode | Default behavior | Agent recommendation |
|---|---|---|
| Read-only | Executes immediately | Safe for discovery, search, reconciliation, and MCP default mode |
| Write | Returns a preview unless --yes is set | Review preview, confirm intent, then execute |
Write families include deposits, intents, vault operations, delegation changes, transfers, and checkout mutations.