Skip to main content
Glama
Poster POS MCP server

Poster POS MCP server

Poster POS MCP server

A remote Model Context Protocol server that lets an AI assistant (Claude, from desktop or phone) read and safely configure Poster POS accounts across a multi-location restaurant/café network.

I built it to run my own 10-location coffee brand from a chat window: check menus, recipes, ingredients, purchases, balances and staff, and make controlled changes, without opening each POS back office.

Why it's interesting

Letting an LLM write to a production POS is risky. The server enforces the safety rules itself, so they don't depend on the model behaving:

Rule

How it's enforced

Account allow-list

Only accounts in CUENTAS (src/config.ts) can be touched, including for reads

Writes default to one account

WRITE_DEFAULT_ACCOUNTS. Any other account needs an explicit permitir_otra_cuenta: true on that call

Dry-run by default

Every write with confirm: false returns the diff, warnings and the exact payload without touching Poster

Apply once, then verify

confirm: true writes once (no retries, to avoid duplicates) and reads back to verify

Audit log

Every dry-run, apply and rejection is logged with who, when, account, method, payload (token-redacted) and result

Secrets never exposed

Tokens come from Google Secret Manager, are redacted from errors and logs, and no tool returns them

No destructive deletes

Items off the menu are hidden, not deleted

Related MCP server: manager-full-mcp

Tools

  • Read: accounts, menu snapshot, product detail, ingredients, prepared items, customer groups, promotions, employees, payment methods, settings, purchases, balances, base catalogs, generic get* passthrough, recent audit.

  • Write (dry-run first): customer groups, ingredients, purchases, dish and category updates, recipes, prepared items, categories, transactions, purchase payments, hide dish.

Stack

TypeScript · Hono · MCP Streamable HTTP (stateless) · OAuth + bearer auth · Google Cloud Run · Secret Manager · Vitest (58 tests)

Run locally

npm install
cp .env.example .env   # PORT, TOKEN_SOURCE, MCP_BEARER_TOKEN, ...
npm run dev
npx vitest run

Deployment notes (in Spanish): DEPLOY_CLOUD_RUN.md. Original Spanish README: README.es.md.

License

MIT © 2026 Roy Camacho Jácome. See LICENSE.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI assistants to query live Toast POS data and generate sales, labor, and cash reports while answering restaurant operations questions, all in a read-only manner.
      33 npm
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      Enables AI assistants to work with the entire Manager.io API surface across any number of businesses, with granular per-resource controls that can block, require confirmation, or allow operations. It surfaces real confirmation dialogs before risky writes and keeps security-sensitive endpoints read-only.
      13
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables AI assistants to manage Google Business Profiles, including updating hours, publishing posts, adding photos, replying to reviews, and reading up to 18 months of performance data. Changes are scoped to owner-approved locations, with sensitive fields locked or held for approval and logged so they can be reviewed and undone.
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables AI assistants to read and safely configure point-of-sale accounts across multi-location restaurant and café networks, with dry-run-by-default writes, account allow-lists, and audit logging.
      MIT