Skip to main content

corral

Round up your projects into herdr workspaces.

corral sets up a herdr workspace for any project under a root directory (~/Code by default). Each workspace gets a utility tab (file manager, shell and git UI by default) and one or more agent tabs, each named for the model and effort it runs: Sonnet•medium, Opus•high, Codex•xhigh. Use the TUI to browse projects and act with single keys, or the CLI (with --json) for scripts and coding agents.

┌ corral 0.1.0 ─────────────────────────── ~/Code ┐
│ ● cEntities        wY  ● Opus xhi  ● Opus hi   develop  │ cruzainet
│ ○ ▸ Archive/ · 5 repos                                   │ ~/Code/cruzainet
│ ○ ▾ cruzainet                             master        │
│ ○   ├─ api                                develop       │ kind     git repo
│ ○   ├─ web-app                            style/visual… │ nested   8 repos below
│ ○   └─ specs                              master        │ branch   master
└ o Open  a Add agent  f Fill  s Stop  x Close WS  / Filter ┘

Install

Requires Python 3.11+ and herdr 0.8+.

uv tool install corral-herdr    # or: pipx install corral-herdr
brew install johnfoland/tap/corral-herdr

The package is corral-herdr on PyPI and in Homebrew; the command it installs is corral. (homebrew/core's corral is the Pony package manager, which also installs a corral command, so the two can't be installed together.)

Use

corral                                  # the TUI
corral up ~/Code/api                    # build the workspace, or focus it if it exists
corral up ~/Code/api --agent opus/high --agent codex/xhigh
corral up --fill                        # add whatever tabs the current project's workspace lacks
corral tab opus/high                    # an agent tab in the workspace you're in
corral tab sonnet --new                 # another one: Sonnet•medium-2
corral stop "Opus•high"                 # by tab label, agent name or pane id
corral close courses --dry-run          # what closing would take with it
corral ls                               # projects, workspaces, agents
corral models                           # the model matrix

Every command takes --json: the result goes to stdout as JSON, progress to stderr. Exit codes: 0 ok, 1 a target failed, 2 usage/config error, 3 herdr not running. --json, --root and --config go before or after the command (corral --root ~/Work ls).

Projects and labels

Every directory in the root is a project. So is every git repo nested up to scan_depth levels inside one (cruzainet/api), along with the plain folders that lead to one (Archive/AyeAI/). A project's workspace is labelled with its path relative to the root, so two nested repos that share a name don't collide. Hidden directories and dependency/build folders (node_modules, vendor, dist, …) are skipped.

Your home directory is a project too: it heads the list as ~, and its workspace is labelled ~ (quote it on the command line: corral close '~').

TUI keys

Key Action
o / enter open: build the workspace, or switch to it
a add an agent tab: pick a model, then an effort (always a new tab)
f / u add any missing tabs / only the utility tab
s stop running agents (pick them)
F force-fill: shows the plan, runs on y
x close the workspace: shows what goes with it, runs on y
→ ← space unfold / fold / toggle the tree
/ g q filter, refresh, quit
, settings

Configure

Press , in the TUI for the settings screen. Its tabs cover the project root (with a folder browser), the agent tabs a new workspace gets, the utility tab's three panes, the model matrix and effort levels, and a few advanced options. Saving writes the config file and keeps its comments and layout. If the root doesn't exist when the TUI starts, the settings screen opens so you can choose one.

From the command line:

corral config init      # writes a commented ~/.config/corral/config.toml
corral config show      # the settings in effect

The file lives at $CORRAL_CONFIG, else $XDG_CONFIG_HOME/corral/config.toml, else ~/.config/corral/config.toml (on macOS too). Every key is optional.

root = "~/Code"                     # also --root / $CORRAL_ROOT
scan_depth = 3
default_agents = ["sonnet/medium"]

[utility]                           # "" = plain shell
top = "yazi"
bottom_left = ""
bottom_right = "lazygit"

[efforts]
claude = ["low", "medium", "high", "xhigh", "max"]

[[models]]                          # any herdr agent kind: claude, codex, gemini, opencode, ...
key = "opus"
tool = "claude"
display = "Opus"
args = ["--model", "opus", "--effort", "{effort}"]

Agent skill

skills/corral/SKILL.md teaches coding agents to drive corral through its --json CLI. The repo is also a Claude Code plugin marketplace:

/plugin marketplace add johnfoland/corral
/plugin install corral@corral

Or copy skills/corral into your agent's skills directory.

Develop

uv sync
uv run pytest
uv run ruff check . && uv run ruff format --check .
uv run ty check

Tests run against an in-memory fake herdr (tests/fake_herdr.py); nothing touches a real herdr session. See CONTRIBUTING.md for commit conventions and the release process, the issues for the roadmap, and docs/decisions.md for settled design questions.

License

MIT

Metadata

Release files for corral-herdr 0.3.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for corral-herdr 0.3.0
File Size Uploaded Provenance
corral_herdr-0.3.0.tar.gz 53.7 kB With provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for corral-herdr 0.3.0
File Interpreter ABI Platform
corral_herdr-0.3.0-py3-none-any.whl Python 3 none any Details

Total release size: 98.2 kB

Release files / corral_herdr-0.3.0.tar.gz

Download URL corral_herdr-0.3.0.tar.gz
Size 53.7 kB
Tags Source
SHA-256 checksum
How to use checksums
d54b23fa0e94b331bf43a76dc0ddcdca1dca24bba0b3dc4c98463f048cd3b1c2
BLAKE2b-256 checksum
How to use checksums
ffd769294bd37f08dac870715ab8c848450c579782bc644b8b5f6dadc4a81fad
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 26, 2026.

Transparency log

Release files / corral_herdr-0.3.0-py3-none-any.whl

Download URL corral_herdr-0.3.0-py3-none-any.whl
Size 44.5 kB
Tags Python 3
SHA-256 checksum
How to use checksums
75a7a9cf6e5ee4b77b0f79a763c11e716da1eba531662b8d41788d562510464c
BLAKE2b-256 checksum
How to use checksums
6abd6081e8f88e80d92a6b311c56aa7db6b9050c8dbc540d2741ecfb7bedf4ce
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 26, 2026.

Transparency log

Security for corral-herdr 0.3.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 2 files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for corral_herdr-0.3.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 26, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page