Exit codes

Kiro CLI returns specific exit codes to indicate operation status. Use these in scripts and CI/CD pipelines to detect success, failures, and specific error conditions.

Exit code reference

Code	Name	Description
0	Success	Command completed successfully
1	Failure	General failure (auth error, invalid args, operation failed)
3	MCP Startup Failure	A server required by `--require-mcp-startup` did not start
4	Requested Agent Not Found	The agent a non-interactive run named was unavailable on the selected agent engine

Requiring MCP servers

By default, MCP server failures are logged as warnings but don't affect the exit code. Use --require-mcp-startup to wait for configured MCP servers before a run begins:

bash

kiro-cli chat --require-mcp-startup --no-interactive "Run task"

If a required server does not start, Kiro CLI exits with code 3. In a non-interactive V3 run, this includes a server failure, an undetermined startup state, or no reported status within 30 seconds.

Requested agent not found

A non-interactive run (--no-interactive, V2 or V3) exits with code 4 when its named agent isn't available on the selected agent engine:

A run names an agent with --agent on either engine. If that agent isn't found, the run exits 4.
On V3, a fresh session also names the chat.defaultAgent setting. If that configured default is unavailable, the run exits 4.
A run with no named agent is not gated. This includes resumed sessions without --agent.

On V3, a named agent that exists only in the V2 format also counts as unavailable. In Kiro CLI 2.37 or later, a stopped run says stopping (exit 4) and, for a V2-format agent, names the kiro-cli agent upgrade <directory> command for that agent's directory. Run it, then retry. See Upgrading agent configs.

Exit code 4 applies regardless of --trust-all-tools.

Other --agent failures do not exit 4. They report the error and exit with code 1.

Scripting examples

Handle different exit codes to take appropriate action in your automation:

Bash script

bash

#!/bin/bash
kiro-cli chat --require-mcp-startup --no-interactive --trust-all-tools "Run analysis"
exit_code=$?

case $exit_code in
    0) echo "Success" ;;
    3) echo "MCP servers failed to start"; exit 1 ;;
    4) echo "Requested agent not found"; exit 1 ;;
    *) echo "Failed with code $exit_code"; exit $exit_code ;;
esac

CI/CD pipeline

yaml

- name: Run Kiro task
  run: |
    kiro-cli chat --require-mcp-startup --no-interactive --trust-all-tools "Analyze code"
  continue-on-error: false

Hook exit codes

Hooks use a separate set of exit codes to control tool execution:

Code	Behavior
0	Hook succeeded
2	(PreToolUse only) Block tool execution; STDERR returned to LLM
Other	Hook failed; STDERR shown as warning

Best practices

Use --require-mcp-startup in CI/CD when your tasks depend on MCP tools
Add verbose logging (-v or -vv) when debugging exit code issues
Check exit codes explicitly rather than relying on implicit shell behavior
Separate MCP failures from general failures to provide better error messages to users

Page updated: November 2, 2026

Built-in tools

Settings