Tokei Agent

tokei-agent

Control your Tokei pre-launch and waitlist campaigns from the command line — and from AI agents like Claude Code and OpenClaw. Wraps the Tokei v1 REST API with JSON-only output, zero runtime dependencies.

Install

npm install -g tokei-agent

or run without installing:

npx tokei-agent --help

Requires Node 22.12+.

Claude Code

Install it as a plugin and the tokei-agent skill comes with it — Claude then knows the whole command surface without being told:

/plugin marketplace add gilesdawe/tokei-agent

/plugin install tokei-agent@tokei

Set TOKEI_API_KEY (see Quick start) and you're done. The CLI itself still runs via npx, so there is nothing else to install.

Quick start

1. Create an API key at tokei.io → Dashboard → Settings → API Keys. Pick read-only unless you need to change things; API access requires an active subscription or lifetime plan.

1. Export it — the syntax differs by shell:

export TOKEI_API_KEY=tokei_k_...          # bash / zsh

set -x TOKEI_API_KEY tokei_k_...          # fish

$env:TOKEI_API_KEY = "tokei_k_..."        # PowerShell

1. Try it:

tokei-agent me                       # verify the key, see plan + API usage

tokei-agent pages:list --status active tokei-agent stats # analytics for one page

Every command prints JSON to stdout with a top-level rate_limit object. Exit codes: 0 success, 1 API/network error, 2 usage error (JSON on stderr).

Interactive output (0.3.1+). When stdout is an interactive terminal, the CLI renders a banner and a human-readable summary instead of raw JSON. When stdout is a pipe, a redirect, CI, or the mcp transport, it prints exactly the JSON it always has — so agents, scripts and MCP clients are unaffected. Set TOKEI_OUTPUT=json to force JSON at a terminal too; NO_COLOR disables colour and animation, and TOKEI_NO_ANIM=1 keeps the colour but stops the movement.

Known issue — exit codes on Node 24 / Windows (fixed in 0.3.0). On 0.2.2 and earlier the CLI could print its correct JSON output and then abort during process exit, corrupting the exit code ($LASTEXITCODE read -1073740791 / 0xC0000409 on success and failure alike). On an affected version, judge a run by the JSON on stdout, not by the exit status — or upgrade. (Historical labelling slip: the 0.3.0 tarball misreported --version as 0.2.2; 0.3.1+ reports correctly.)

Commands

Read (any key):

Write (needs a read+write key):

Write commands take simple fields as flags and full/nested bodies via --data ' ' or --data @file.json (flags win on conflict). Run tokei-agent --help for every flag, or see SKILL.md — the agent-oriented reference bundled in this package, with worked examples and error-handling guidance.

tokei-agent pages:clone --title "Spring Launch Waitlist" --source <promotionId>

tokei-agent pages:update --end-date 2026-09-01T00:00:00Z tokei-agent media:upload ./hero.png tokei-agent pages:update --image-video tokei-agent actions:catalog --type twitter_follow # see what a type accepts tokei-agent pages:update \ --data '{"entry_methods":[{"actionType":"tiktok_follow","label":"Follow us on TikTok","points":3,"config":{"username":"tokei"}}]}'

MCP server

The package doubles as a local MCP server (stdio): every command above becomes an MCP tool (pages_list, pages_update, stats, …) for Claude Code, Claude Desktop, and other MCP clients.

claude mcp add tokei --env TOKEI_API_KEY=tokei_k_... -- npx -y tokei-agent mcp

Or in JSON config:

{

"mcpServers": { "tokei": { "command": "npx", "args": ["-y", "tokei-agent", "mcp"], "env": { "TOKEI_API_KEY": "tokei_k_..." } } } }

Environment

Agents, human in the loop

Tokei is built so agents draft and humans approve: give monitoring agents a read-only key, reserve read+write keys for agents that genuinely need to change things, and set key expiry. New webhook subscriptions created via the API trigger a security notification to the account owner.

Privacy Policy

Full policy: https://tokei.io/privacy

Data collection practices

tokei-agent is a thin client for the Tokei v1 REST API. It collects no data of its own: there is no telemetry, no analytics, no crash reporting, and no phone-home of any kind.

It reads exactly two inputs from your environment — TOKEI_API_KEY and the optional TOKEI_API_URL — plus the arguments you pass on the command line (or the arguments an MCP client passes to a tool call). For media:upload it also reads the bytes of the local file you name.

Usage and storage

Your API key is used solely as an Authorization: Bearer header on requests to your Tokei account. Command arguments become the request path, query string, or JSON body. Nothing is written to disk: the CLI creates no config file, cache, credential store, or log file, and holds nothing after the process exits.

API responses — which can include entrant email addresses, survey answers, and analytics — are printed as JSON to stdout. From that point they are handled by whatever invoked the CLI (your shell, your scripts, or your AI agent and its conversation history). Treat that output as the personal data it is.

Third-party sharing

Nothing is sold or shared with third parties. Data is transmitted only to:

• Tokei (https://tokei.io, or the host you set in TOKEI_API_URL) — every command, to serve your request against your own account.
• Tokei's object storage provider — media:upload only. The CLI sends the file to the Tokei API, which stores it with the provider; the CLI never contacts the provider itself. The file becomes a public asset on your promotion page.

The CLI contacts no other host.

Data retention

The CLI retains nothing. Data held in your Tokei account — promotions, entries, survey responses — is retained under the Tokei privacy policy above, and you can request access or deletion there. API keys are created, scoped, expired, and revoked by you at Dashboard → Settings → API Keys; revoking a key immediately stops all access through it.

Contact information

Tokei — https://tokei.io/privacy — support@tokei.io

Docs

• API reference: https://tokei.io/docs/api
• OpenAPI spec: https://tokei.io/openapi.json
• Agent skill reference: SKILL.md

License

MIT