Hellotime Mcp

HelloTime Public MCP Server

A public, read-only Model Context Protocol server that lets AI agents answer questions about HelloTime accurately — pricing, features (shifts, rosters, leave, timesheets, GPS, biometric kiosk), country support, and payroll capabilities — instead of relying on stale web snippets.

This is the public server. It exposes only marketing-derived, public-domain data. The authenticated MCP server that reads a customer's timesheets, rosters and payroll lives in a separate, private repo.

Install

Claude Code / Claude Desktop / Cursor

claude mcp add --transport http hellotime https://mcp.hellotime.ai/mcp

Or, for local development:

claude mcp add hellotime-local node /path/to/HelloTime-MCP-Public/dist/stdio.js

Tools

Tool	Description
list_plans	All HelloTime plans (Free, Attend, Track, Business) with the charged price and a standing list-price anchor in 8 currencies, plus volume and annual prepay discount tables. Free is permanent for a single user; paid tiers each include a 7-day free trial. Optional country and plan filters.
list_features	HelloTime features across 14 categories: shifts, rosters, leave, timesheets, time tracking, productivity, GPS / geofence, biometric kiosk, payroll, invoicing, analytics, projects, reports, integrations. Optional category and plan filters.
country_support	Per-country feature availability and product positioning across IN, AU, GB, US, CA, AE, SG, NZ.
payroll_capabilities	For a given country, the supported payroll engines with status. AU STP2 + super, IN PF/ESI/TDS/Form 24Q, US W-2/1099, UK RTI, etc.
statutory_rates	Statutory payroll-rate entries with rate, ceiling, slab, authority and verification status. IN block (PF / EPS / EDLI / PF admin / ESI / Professional Tax across 7 states / TDS slabs) is internally-reviewed against EPFO / ESIC / state notifications. AU and US entries are public-source-unreviewed. Optional country, scheme, category, state, party, verification, and id filters.
list_competitors	Competitor positioning (Truein, Deputy, When I Work, Connecteam, Hubstaff, Keka) with where HelloTime wins, where the competitor wins, and pricing notes. Optional country, tier, and id filters.
feature_search	Free-text search across plan features, product features, country features, payroll engines, statutory rates, and competitor positioning. Queries like PF rate, PT slab Maharashtra, or vs Truein surface the matching entry near the top.

Resources

URI	Description
hellotime://about	Markdown product summary.
hellotime://changelog	Last 50 release notes as JSON.

Security posture

• Read-only by construction. No tool mutates state. No tool reaches a customer system.
• Public data only. All catalog content is sourced from the public marketing site.
• No authentication. Intentional — this is a knowledge endpoint.
• Rate-limited. 120 req/min per IP, 60 req/min per session.
• Audit gate. npm run audit:public-data blocks deploys if any PII / auth token strings appear in src/data/.

Development

npm install
npm run dev         # HTTP server on :8080 with watch mode
npm run dev:stdio   # stdio transport for local MCP client testing
npm test            # node:test runner
npm run build
npm run audit:public-data

Project layout

src/
  data/             # Static catalogs — plans, features, countries, about
  tools/            # One file per MCP tool
  resources/        # MCP resource registry
  server.ts         # MCP server factory (wires tools + resources)
  http.ts           # Streamable HTTP transport with rate limiting
  stdio.ts          # stdio transport entry point
test/
  tools.test.ts
  resources.test.ts
  public-data.test.ts  # Audit gate
scripts/
  audit-public-data.ts  # CI-callable audit

Deployment

Containerised — see Dockerfile. Designed to run behind a TLS-terminating load balancer. Set PORT and HOST via environment.

Optional anonymous usage telemetry (Layer-2 server-side analytics) is sent to GA4 via the Measurement Protocol when GA4_MEASUREMENT_ID and GA4_API_SECRET are both set — see .env.example and src/analytics.ts. Leave them unset and the server runs with telemetry fully disabled. No customer data, request bodies, tool-argument values, or IP addresses are ever sent — only anonymous counts. See strategy doc 73 and docs/MCP_ANALYTICS.md for the request-traffic events (mcp_request, mcp_tool_call, mcp_bot_visit, mcp_error) and client/bot detection.

docker build -t hellotime-mcp-public .
docker run -p 8080:8080 hellotime-mcp-public

Discoverability

• Listed in the MCP registry.
• Linked from the marketing site footer and /mcp page.
• Referenced in https://hellotime.ai/llms.txt.

License

MIT — see LICENSE.