Shell scripts · CI jobs · coding agents

Automate Huly from the command line

Huly CLI exposes the complete shared Huly operation surface through predictable commands, structured JSON, file-backed inputs, and process-friendly failures.

Make JSON the script boundary

Add --json for machine-readable stdout. When an operation also emits warnings, the CLI preserves them alongside the result.

# Select fields from a project listing
huly projects list --json | jq '.projects[] | {identifier, name}'

# Keep a complete result for the next automation step
huly issues get PROJ 123 --json > ./issue.json

Keep complex inputs in files

--input-file accepts a complete JSON operation input. This avoids brittle shell quoting for arrays and nested values.

huly issues create \
  --input-file ./new-issue.json \
  --json

Inputs merge in command-line order. Named positionals and explicit flags override JSON values, and file-backed field flags have the final precedence.

Use process semantics

stdout for results

Structured results stay on stdout so they can be piped, redirected, or parsed.

stderr for failures

Typed operation and input failures are rendered on stderr with a non-zero exit code.

`--yes` for intent

Pass explicit confirmation for deletion, access changes, and other consequential operations.

Run without a global install

Use @latest for ad hoc jobs. For reproducible automation, replace it with the exact published version your workflow has tested.

pnpm dlx @firfi/huly-cli@latest projects list --json

Choose the right agent integration

The CLI works well when an agent can run subprocesses. Huly MCP is purpose-built for agents that discover and invoke tools through Model Context Protocol.

Compare CLI and MCP Browse examples