Six open-source, opinionated agents that turn the most common project-manager tasks into a sentence you type instead of an hour you click through. Works on Claude Code and Gemini CLI.
Find the bottleneck. Reassign the work. Clone the engagement. Scan for risk. Prep your standup. Draft the update. All from your terminal, all using the Smartsheet MCP tools, all installable in sixty seconds.
Six local agents for Claude Code and Gemini CLI, purpose-built against the Smartsheet MCP server. Each one is a single markdown file — under .claude/agents/ for Claude Code, .gemini/agents/ for Gemini CLI. They activate automatically when you ask the right question, and they come pre-loaded with twenty years of Smartsheet behavior baked in.
| Stage | Power Tool | The prompt | What it typically replaces |
|---|---|---|---|
| READ | bottleneck-scanner |
"Who's the bottleneck across my active projects?" | Opening sheets, counting tasks by owner, cross-checking thread context |
| WRITE | reassignment-helper |
"Reassign everything from Alex to Jordan." | Filtering each sheet, clicking each row, checking access |
| CREATE | engagement-cloner |
"Clone this project sheet for a new engagement." | Finding a template, copying, clearing data, renaming, re-sharing |
| SCAN | risk-scanner |
"What's at risk across my active projects?" | Manually checking due dates, blocked rows, and unassigned tasks |
| BRIEF | standup-prep |
"Prep my standup." | Pulling up sheets before a meeting to remember what to say |
| DRAFT | status-comms-writer |
"Draft a Slack update from this risk scan." | Translating project data into a message someone else will actually read |
These aren't just prompts. Each Power Tool is a local agent on your machine that orchestrates deep calls into our platform — reading discussion threads, writing to the work graph, composing against Smartsheet's production MCP tools, all operating within your Smartsheet session. Local intelligence on your terminal, commanding twenty years of Smartsheet behavior.
Two arcs, one philosophy.
Arc 1: Read → Write → Create is the trust-building sequence. Start with the question you'd be embarrassed to ask a human — read-only, nothing changes, you just want to know. Move to the change you used to do by hand — previewed, confirmed, batched. End at the creation task you're tired of doing at the start of every engagement.
Arc 2: Scan → Brief → Draft is the daily rhythm. Knowing what's at risk informs what's worth raising. Knowing what's worth raising informs what to communicate. Each stage narrows the signal so the next stage doesn't have to.
Most teams try to lead with the most impressive demo. Most teams burn out on AI adoption because the impressive demo was a write operation that made a mistake in week two. These arcs are ordered the way real trust gets built — start with reads, earn writes, then make it a habit.
# 1. Clone this pack
git clone https://github.com/smartsheet/cli-agent-power-tools
cd cli-agent-power-tools
# 2. Get your Smartsheet Personal Access Token
# Go to Account > Personal Settings > API Access in Smartsheet to generate oneClaude Code:
# Install Claude Code
npm install -g @anthropic-ai/claude-code
# Set up Smartsheet MCP connection (choose one option):
# Option A: Run the automated setup script
./smartsheet_mcp_setup.sh
# Option B: Manual setup following Smartsheet's documentation
# https://developers.smartsheet.com/ai-mcp/smartsheet/install-the-smartsheet-mcp-server/connect-claude-code
# Start Claude Code — Power Tools load automatically from .claude/agents/
claudeGemini CLI:
# Install Gemini CLI
npm install -g @google/gemini-cli
# Set up Smartsheet MCP connection
./gemini_mcp_setup.sh
# Start Gemini CLI — Power Tools load automatically from .gemini/agents/
geminiThen just ask:
> Who's the bottleneck across my active projects?
Power Tools route automatically based on what you ask. You can also invoke them explicitly: use bottleneck-scanner on the Healthcare practice. In Gemini CLI, you can also force-invoke with @bottleneck-scanner.
Worth making this explicit, because the words look similar.
Smartsheet's platform sub-agents are the production-grade, credit-bearing intelligence layer inside SmartAssist — Risk Analysis, Dependency Detection, Executive Summary, Resource Optimizer, and others. They run on Smartsheet's infrastructure, use proprietary execution data across 100K+ organizations, and are what customers are paying for when they use the platform's AI features.
CLI Agent Power Tools are something different. They're personal prompt configurations that live on your laptop, run in your Claude Code session, and use only the read/write/create MCP tools our server already exposes publicly. They don't invoke platform sub-agents. They don't carry Smartsheet's proprietary intelligence. They're the "get more out of the MCP tools you already have" toolkit.
Think of it this way: platform sub-agents are the industrial machinery. Power Tools are the sharp hand tools you keep in your desk drawer. Both useful, both Smartsheet-made, different jobs.
cli-agent-power-tools/
├── .claude/
│ └── agents/
│ ├── bottleneck-scanner.md ← READ
│ ├── reassignment-helper.md ← WRITE
│ ├── engagement-cloner.md ← CREATE
│ ├── risk-scanner.md ← SCAN
│ ├── standup-prep.md ← BRIEF
│ └── status-comms-writer.md ← DRAFT
├── .gemini/
│ └── agents/
│ ├── bottleneck-scanner.md ← READ
│ ├── reassignment-helper.md ← WRITE
│ ├── engagement-cloner.md ← CREATE
│ ├── risk-scanner.md ← SCAN
│ ├── standup-prep.md ← BRIEF
│ └── status-comms-writer.md ← DRAFT
├── .mcp.json # Shared team MCP config
├── smartsheet_mcp_setup.sh # Claude Code MCP setup
├── gemini_mcp_setup.sh # Gemini CLI MCP setup
├── LICENSE # MIT
├── CONTRIBUTING.md # How to submit new Power Tools
├── USAGE.md # Invocation, task automation, chaining
└── README.md # This file
One git pull gets your whole team the same setup.
This is v1.1. If these find their people, we'll ship new expansion packs. Here's what we're thinking about — subject to change, open to contribution:
- Governance pack —
data-quality-auditor,dropdown-standardizer,stale-cleanup,permission-sweep. For the PMO that owns sheet hygiene at scale. - Setup pack —
workspace-organizer,template-converter,starter-sharer. For the first 30 days of a new team or customer onboarding.
Pull requests welcome. See CONTRIBUTING.md.
Everyone in enterprise software this quarter is gluing a chatbot onto a shallow product and calling it AI. Talk to your todo list. Ask your spreadsheet a question. Cute demos; not much underneath.
Our MCP isn't reaching into a todo list. It's reaching into twenty years of workflow engine, a full dependency graph, row-level discussion threads, workspace hierarchy, formula resolution, cross-sheet references, permission model, and governance built before "AI governance" was a phrase anyone said out loud.
Put a terminal on top of a calendar app — you get a gimmick. Put a terminal on top of that — and a project manager can run a portfolio from it.
The CLI isn't the point. The work graph is the point. Local agents are the mechanism that lets you compose against twenty years of operational depth, at the speed of a sentence.
For the technically curious: CLI Agent Power Tools are implemented as local sub-agents — markdown files with YAML frontmatter that both Claude Code and Gemini CLI support natively. Claude Code loads them from .claude/agents/; Gemini CLI loads them from .gemini/agents/. One clone, both runtimes. We call them Power Tools in our copy to keep them clearly distinct from Smartsheet's platform sub-agents, which are a separate product concept. The mechanism is Anthropic's and Google's; the domain expertise baked in is ours.
MIT. Take it, fork it, ship derivatives. See LICENSE.
Built by the Smartsheet AI Platform team. Maintained by Drew Garner, SVP of AI & Platform Strategy.