agent-toolkit banner

agent-toolkit-cli

The official CLI for agent-toolkit — install skills, agents, loops, and MCP configs across every major AI coding assistant.

skills agents loops

Monorepo · Docs · Installation · CLI surfaces · Changelog

What is this package?

agent-toolkit-cli is the PyPI adapter for the agent-toolkit monorepo (ADR-021). Platform wheels bundle the native V binary; agent-toolkit / agent-toolkit-cli are a thin trampoline (same idea as the npm package). There is no Python CLI fallback — see python-fallback.md.

It also ships:

bundled skills, agents, loops, profiles, packs, and MCP templates
install / doctor / sync workflows that deploy the right artifacts per AI tool

One CLI. Any coding assistant. Zero duplication.

uvx --from agent-toolkit-cli agent-toolkit install
agent-toolkit doctor

Quick start

Recommended: uv — no permanent install required.

# One-shot (preferred)
uvx --from agent-toolkit-cli agent-toolkit install
uvx --from agent-toolkit-cli agent-toolkit doctor

# Or install the CLI persistently
uv tool install agent-toolkit-cli
agent-toolkit install    # auto-detects Claude, Cursor, OpenCode, Windsurf, Pi, Copilot
agent-toolkit doctor

Other installers

pip install agent-toolkit-cli
pipx install agent-toolkit-cli

# System packages (release notify → dedicated taps)
brew tap ulises-jeremias/homebrew-tap && brew install agent-toolkit
yay -S agent-toolkit-bin   # Arch Linux (AUR) — GitHub Release V binary; not the Python AUR package
docker pull ghcr.io/ulises-jeremias/agent-toolkit

Full walkthrough: docs/INSTALLATION.md

Key concepts

🛠️ Skills _{Reusable capability units (SKILL.md) that teach an agent how to do a job — delivery, forge CLIs, design, data, ops.} _{80 skills across domains. Browse with agent-toolkit inventory or agent-toolkit skills list.}	🤖 Agents _{Personas that constrain how the AI works — review, plan, architect, fix CI — without rewriting prompts each time.} _{17 personas, compiled into each target's native format via install.}
🔄 Loops _{Recurring agentic workflows with mutation-safety tiers (L1 observe → L2 controlled → L3 high autonomy).} _{agent-toolkit loop run daily-triage · 10 bundled templates}	🔗 MCP _{Provider registry + ready templates (GitHub, Slack, Notion, Linear, Figma, ClickUp) emitted into target-native configs.} _{agent-toolkit mcp list · mcp setup <provider> · mcp doctor}

Supported tools

Tool	What `install` deploys
Claude Code	Plugin manifest, skill references, settings
Cursor	Marketplace plugins + `.mdc` rules (IDE + Agent CLI)
OpenCode	System prompt overlays, agent configs
GitHub Copilot	`copilot-instructions.md` with domain selection
Windsurf	Rules and memory files via Cascade (`~/.codeium/windsurf/`)
Pi Agent	Skills and loop templates in Pi's native format
Muse Code	Agent Skills under `~/.config/muse/skills/`

# Target specific tools
agent-toolkit install --tools claude-code,cursor
agent-toolkit install --dry-run
agent-toolkit install --force

Command reference

Everyday consumer commands vs advanced workstation / harness commands. Details: docs/CLI_SURFACES.md

Consumer

Command	Purpose
`install`	Install profiles for detected or selected AI tools
`update`	Refresh installed profiles from latest toolkit data
`uninstall`	Remove toolkit-owned files using install receipts
`doctor`	Check toolkit data and tool availability
`diff`	Show changes vs currently installed plugin bundles
`skills`	Sync, list, and validate skills
`mcp`	MCP provider setup, health, doctor, uninstall
`plugin`	Plugin bundle sync and check

agent-toolkit skills list
agent-toolkit skills sync --tools claude-code,cursor
agent-toolkit mcp setup github
agent-toolkit doctor --fix

Advanced

Still on the same binary — start here when running an agentic-harness-style workspace:

Command	Purpose
`loop`	Loop engineering: init, run, status, audit, cost, schedule, sync
`workspace`	Workspace scaffolding: init, context, personas, packs
`memory`	Knowledge base: add, search, inject, review, todo
`project`	Project index: clone, list, add, remove, scan
`devcompanion`	Background job queue (`dc` alias)
`insights`	Local usage analytics (OpenCode, Cursor, Claude)
`inventory`	List skills, agents, and products
`matrix`	Platform capability matrix
`build` / `release`	Compile / release artifacts (maintainer)

# Loops (mutation-safety tiers L1 → L3)
agent-toolkit loop init daily-triage
agent-toolkit loop run daily-triage
agent-toolkit loop status
agent-toolkit loop audit daily-triage

# Workspace harness
agent-toolkit workspace init --dir ~/.agentic-harness
agent-toolkit memory search "topic"
agent-toolkit project clone owner/my-repo
agent-toolkit inventory

Tier	Posture	Examples
L1	Observe / propose (report-only)	`oss-daily-briefing`, `daily-triage`, `oss-triage`
L2	Controlled mutations (PR-gated)	`ci-sweeper`, `dep-sweeper`, `pr-babysitter`
L3	High autonomy (allowlist)	`oss-pr-monitor`

Optional dependencies

# YAML-backed pack / catalog features
uv tool install "agent-toolkit-cli[yaml]"

# Everything optional
uv tool install "agent-toolkit-cli[all]"

Extra	Provides
`yaml`	`PyYAML` for pack overlays and richer catalog parsing
`all`	Currently same as `yaml`
`dev`	`pytest`, `ruff` (contributors)

Upgrade & uninstall

uv tool upgrade agent-toolkit-cli
agent-toolkit update          # refresh deployed profiles
agent-toolkit uninstall       # remove toolkit-owned files via receipts

uv tool uninstall agent-toolkit-cli
# or: pip uninstall agent-toolkit-cli

→ docs/UNINSTALL.md

Ecosystem

agent-toolkit is the capability distribution layer (L1.5) in a personal DX stack:

Layer	Repo	Role
L1	agentic-workstation	Machine provisioning — chezmoi, shell, packages, LLM policy
L1.5	agent-toolkit (this CLI)	Skills, loops, profiles, MCP
L3	agentic-harness	AI workspace scaffold — memory, personas, multi-repo orchestration

Documentation

Guide	Description
README (monorepo)	Full product overview, skills, agents, packs
INSTALLATION.md	Install paths per tool
CLI_SURFACES.md	Consumer vs advanced commands
LOOPS.md	Loop engineering reference
TRUST.md	Trust boundaries and supply chain
COMPATIBILITY.md	Python / tool matrix
Wiki	Extended reference

Canonical sources live in the agent-toolkit monorepo. This package is the installable distribution of that source of truth.

Requirements

Python 3.10+
Optional: uv for uvx / uv tool workflows
At least one supported AI coding assistant (for install to deploy profiles)

Contributing

Issues and PRs welcome on the monorepo:

git clone https://github.com/ulises-jeremias/agent-toolkit
cd agent-toolkit
uv sync --project packages/pypi/agent-toolkit-cli --all-extras
uv run --project packages/pypi/agent-toolkit-cli --directory . agent-toolkit --help

MIT © ulises-jeremias · PyPI · npm · AUR · GitHub

Release files for agent-toolkit-cli 1.26.0

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

Source distribution (sdist)

Source distribution for agent-toolkit-cli 1.26.0
File	Size	Uploaded
agent_toolkit_cli-1.26.0.tar.gz	1.1 MB	Aug 31, 2026	Details

Built distributions (wheels)

Table of built distributions (wheels) for agent-toolkit-cli 1.26.0
File	Interpreter	ABI	Platform	Reset
agent_toolkit_cli-1.26.0-py3-none-win_amd64.whl 7.4 MB Aug 31, 2026	Python 3	none	Windows x86-64	Details
agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_x86_64.whl 7.4 MB Aug 31, 2026	Python 3	none	Linux glibc 2.38+ x86-64	Details
agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_aarch64.whl 7.8 MB Aug 31, 2026	Python 3	none	Linux glibc 2.38+ ARM64	Details
agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_x86_64.whl 6.1 MB Aug 31, 2026	Python 3	none	macOS 11.0+ x86-64	Details
agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_arm64.whl 6.1 MB Aug 31, 2026	Python 3	none	macOS 11.0+ ARM64	Details

Total release size: 36.0 MB

Release files / agent_toolkit_cli-1.26.0.tar.gz

Download URL	agent_toolkit_cli-1.26.0.tar.gz
Size	1.1 MB
Tags	Source
SHA-256 checksum How to use checksums	`d75f0efc38f2341bff50367bf72d6df36d75c61ce5081520277625669e5a6125`
BLAKE2b-256 checksum How to use checksums	`e53825c0e18fad88ce4ea83d0ba62ed82bfe268985684d89b10ef737c8d4336d`
Upload date	Aug 31, 2026
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.

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/ulises-jeremias/agent-toolkit
Publishing commit: github.com/ulises-jeremias/agent-toolkit/tree/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4

Workflow

Publishing configuration: github.com/ulises-jeremias/agent-toolkit/blob/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4/.github/workflows/release.yml
Publishing logs: github.com/ulises-jeremias/agent-toolkit/actions/runs/33362952368/attempts/1

Artifact

Subject: agent_toolkit_cli-1.26.0.tar.gz
SHA-256 checksum: d75f0efc38f2341bff50367bf72d6df36d75c61ce5081520277625669e5a6125 Verifying attestations

Release files / agent_toolkit_cli-1.26.0-py3-none-win_amd64.whl

Download URL	agent_toolkit_cli-1.26.0-py3-none-win_amd64.whl
Size	7.4 MB
Tags	Python 3 Windows x86-64
SHA-256 checksum How to use checksums	`79f709220c0df9f505b7fc652e1f07c8765a5798e0ab641c92aa6f2c999ff1bb`
BLAKE2b-256 checksum How to use checksums	`bee4b3e0b0036c0b22b9e95ed6bc361e650667a1d19094ff944ce060eb6bef5d`
Upload date	Aug 31, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/ulises-jeremias/agent-toolkit
Publishing commit: github.com/ulises-jeremias/agent-toolkit/tree/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4

Workflow

Publishing configuration: github.com/ulises-jeremias/agent-toolkit/blob/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4/.github/workflows/release.yml
Publishing logs: github.com/ulises-jeremias/agent-toolkit/actions/runs/33362952368/attempts/1

Artifact

Subject: agent_toolkit_cli-1.26.0-py3-none-win_amd64.whl
SHA-256 checksum: 79f709220c0df9f505b7fc652e1f07c8765a5798e0ab641c92aa6f2c999ff1bb Verifying attestations

Release files / agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_x86_64.whl

Download URL	agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_x86_64.whl
Size	7.4 MB
Tags	Linux glibc 2.38+ x86-64 Python 3
SHA-256 checksum How to use checksums	`e7273b77f97dcc7dc0ebb692d2e4709194fc36969513e32f9c450fc85c92a1e6`
BLAKE2b-256 checksum How to use checksums	`9e359eaf8267e223ff04cfaeb6c30e8234f4a1ad9e8e3b653a24de5ca71513dd`
Upload date	Aug 31, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/ulises-jeremias/agent-toolkit
Publishing commit: github.com/ulises-jeremias/agent-toolkit/tree/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4

Workflow

Publishing configuration: github.com/ulises-jeremias/agent-toolkit/blob/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4/.github/workflows/release.yml
Publishing logs: github.com/ulises-jeremias/agent-toolkit/actions/runs/33362952368/attempts/1

Artifact

Subject: agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_x86_64.whl
SHA-256 checksum: e7273b77f97dcc7dc0ebb692d2e4709194fc36969513e32f9c450fc85c92a1e6 Verifying attestations

Release files / agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_aarch64.whl

Download URL	agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_aarch64.whl
Size	7.8 MB
Tags	Linux glibc 2.38+ ARM64 Python 3
SHA-256 checksum How to use checksums	`e123146887fab4efd358977f016797f3ee02af72f44a89b1725d61ea85c1e843`
BLAKE2b-256 checksum How to use checksums	`dd7fb2fafac40df4daaf86ed7a2537d678a2f2a652da17a3592152ab9fb08f9f`
Upload date	Aug 31, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/ulises-jeremias/agent-toolkit
Publishing commit: github.com/ulises-jeremias/agent-toolkit/tree/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4

Workflow

Publishing configuration: github.com/ulises-jeremias/agent-toolkit/blob/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4/.github/workflows/release.yml
Publishing logs: github.com/ulises-jeremias/agent-toolkit/actions/runs/33362952368/attempts/1

Artifact

Subject: agent_toolkit_cli-1.26.0-py3-none-manylinux_2_38_aarch64.whl
SHA-256 checksum: e123146887fab4efd358977f016797f3ee02af72f44a89b1725d61ea85c1e843 Verifying attestations

Release files / agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_x86_64.whl

Download URL	agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_x86_64.whl
Size	6.1 MB
Tags	Python 3 macOS 11.0+ x86-64
SHA-256 checksum How to use checksums	`fcde9cdeaf5cebfe00b987a99456c0de8f3308d8de75287b7b0be20890817f45`
BLAKE2b-256 checksum How to use checksums	`7eb2510fba77896b3b9a2cf9869cf780866b3dff06e015cfa6dbd5d96240e276`
Upload date	Aug 31, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/ulises-jeremias/agent-toolkit
Publishing commit: github.com/ulises-jeremias/agent-toolkit/tree/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4

Workflow

Publishing configuration: github.com/ulises-jeremias/agent-toolkit/blob/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4/.github/workflows/release.yml
Publishing logs: github.com/ulises-jeremias/agent-toolkit/actions/runs/33362952368/attempts/1

Artifact

Subject: agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_x86_64.whl
SHA-256 checksum: fcde9cdeaf5cebfe00b987a99456c0de8f3308d8de75287b7b0be20890817f45 Verifying attestations

Release files / agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_arm64.whl

Download URL	agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_arm64.whl
Size	6.1 MB
Tags	Python 3 macOS 11.0+ ARM64
SHA-256 checksum How to use checksums	`11e7924fe6ca2c5330f18664957c62d28cfe467174744796e0e10035928eb879`
BLAKE2b-256 checksum How to use checksums	`251ecacdb48959fb993f501eb8d00f0efaf2e720548ccc21f1f55a9fba56d6bb`
Upload date	Aug 31, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/ulises-jeremias/agent-toolkit
Publishing commit: github.com/ulises-jeremias/agent-toolkit/tree/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4

Workflow

Publishing configuration: github.com/ulises-jeremias/agent-toolkit/blob/f1ac85a5f88bec8d4a58d5fd90388b2d89e6d9d4/.github/workflows/release.yml
Publishing logs: github.com/ulises-jeremias/agent-toolkit/actions/runs/33362952368/attempts/1

Artifact

Subject: agent_toolkit_cli-1.26.0-py3-none-macosx_11_0_arm64.whl
SHA-256 checksum: 11e7924fe6ca2c5330f18664957c62d28cfe467174744796e0e10035928eb879 Verifying attestations