Shakers CLI and MCP
Shakers for freelance talent, from your terminal or your AI assistant. It scans your AI-tool setup on your own machine and shows how you work with AI. It also lets you sign up on Shakers, keep your profile up to date, find projects and ask Alma, the Shakers AI assistant.
Install
Requires Node 18 or later.
Command line (npm)
npm i -g shakers-cli
Or run it without installing: npx -y shakers-cli <command>.
Command line (install script)
curl -fsSL https://raw.githubusercontent.com/devShakers/ai-usage-evaluator/release/install.sh | bash
Uninstall with the same line ending in | bash -s -- --uninstall.
Claude Code plugin
/plugin marketplace add devShakers/ai-usage-evaluator
/plugin install shakers@shakers-mcp
The plugin adds the Shakers MCP server and these slash-commands: onboarding, ai_usage,
my_profile, find_projects, certify and ask.
Claude Desktop
Download the .mcpb file from the latest
GitHub Release and open it
with Claude Desktop.
Main commands
shakers ai-usage # scan this project and machine, show your AI-usage tier (no account needed)
shakers register # create your Shakers account and profile
shakers login # sign in (email or Google)
shakers logout # forget the stored session
shakers certify # certify a dimension of your role through an interview
shakers report # build and open your shareable HTML report
shakers me # your profile summary
shakers find-projects # projects and positions available to you
shakers applications # projects you applied to, and their status
shakers availability # view or set your availability
shakers rate # view or set your rate
shakers alma # chat with Alma
shakers mcp # run the MCP server for an AI desktop app
Run shakers with no command to see every command.
Privacy
- The
ai-usagescan runs on your machine and its report is shown there first. - Saving it to Shakers is opt-in. Every time, before anything is uploaded, you see what
would be sent and what Shakers does with it, and it is uploaded only if you say yes. You
can leave out agents and text excerpts there.
--consent-revokestops future uploads from this computer; it does not delete what was already uploaded. To ask for that, write to talent@shakersworks.com. - What is sent is mostly derived data (counts, categories, tier, detected technologies, sessions, active days, commits and lines changed). Literal text travels too, and the review shows how much: up to 40 fragments of your messages to AI assistants and 20 commands they ran (200 characters at most each), and the definition (up to 2,000 characters) and short description of each of your AI agents, all after removing anything that looks like a secret, a path or an email. The fragments come from the conversations in the repositories analysed, or from every conversation on this computer when you analyse the whole machine (registration always does). Never your code, your full conversations, the names of your repositories, environment variables or credentials.
- An AI model from an external provider (OpenAI, Google or Anthropic) analyses that text automatically for Shakers to compute your evaluation, and then it is discarded. Only the result is stored: levels, scores, counts and the names of your tools and agents.
- Do not run it on a machine or with conversations that hold information you are not allowed to share: client, employer or third-party data. Secret removal is automatic and best effort; names or business details you wrote can remain in the fragments.
- Certification sends secret-scrubbed samples of your code and your interview answers to Shakers to assess them.
- Read the full policy at https://www.shakersworks.com/en/politica-de-privacidad.
License
Proprietary, source available. See LICENSE.
