Projects

AI-friendly Odoo integration — progressive knowledge modules, TypeScript client, and CLI for working with Odoo v17 ERP. What It Does Teaches AI agents how to work with Odoo through progressive, CI-tested skill modules. Instead of dumping 12,000 pages of Odoo docs into context, agents load only what they need: connection → domains → CRUD → domain-specific (mail, accounting, timesheets). Every code example in the skills is extracted and tested against a real Odoo v17 instance in CI. No doc rot, no hallucinated APIs. ...

Unified TypeScript library and CLIs for Gmail, Drive, Calendar, and Tasks — designed for AI agents. What It Does Wraps Google’s official APIs in a single library with four agent-friendly CLIs: go-gmail: Read, search, send email; manage drafts; handle threads go-drive: Upload, download, search files; manage folders go-calendar: Create events, manage calendars, query availability go-tasks: Create, update, list tasks across task lists Two-phase OAuth works in headless environments (auth in browser, token to server). Pagination, encoding quirks, and token refresh handled transparently. ...

Async Python SDK for Odoo’s JSON-RPC API — typed, tested, ready. Built with httpx, dataclasses, and mypy --strict. What It Does A Python client for Odoo with full type annotations, 8 domain services, and testcontainers integration for testing against real Odoo 17, 18, and 19 instances. Packages Package Install Description godoo uv add godoo Core client + 8 domain services godoo-testcontainers uv add godoo-testcontainers Docker-based Odoo for integration testing godoo-introspection uv add godoo-introspection Schema discovery + codegen (coming soon) Quick Start import asyncio from godoo import OdooClient async def main(): async with OdooClient() as odoo: partners = await odoo.records.search_read( "res.partner", [("is_company", "=", True)], limit=5 ) for p in partners: print(p["name"]) asyncio.run(main()) What’s Inside 8 domain services: records, mail/chatter, modules, accounting, timesheets, attendance, contracts, introspection Testcontainers: Spin up real Odoo instances in Docker for integration tests Full type annotations: mypy --strict clean Multi-version: Tested against Odoo v17, v18, and v19 Repository · Documentation · PyPI

An advanced alternative to n8n’s built-in Odoo node, built on odoo-toolbox. Why Replace the Built-in Node? The built-in n8n Odoo node supports basic CRUD on three hardcoded models plus a generic “Custom Resource.” This node replaces that with a unified, model-agnostic approach and adds capabilities that don’t exist in the built-in node. What’s Different Built-in This node CRUD on any model Via Custom Resource ✓ First-class Method calls (action_confirm, etc.) — ✓ Chatter (internal notes & messages) — ✓ Properties (Odoo 17+ dynamic fields) — ✓ Schema introspection — ✓ URL generation (backend, portal) — ✓ Polling trigger — ✓ Domain filters on all operations Custom only ✓ Context support (lang, company) — ✓ Typed error handling Generic Auth / Validation / Access / Missing Install In n8n: Settings → Community Nodes → enter @marcfargas/n8n-nodes-odoo → Install. ...

A command-line interface for Odoo ERP. Query records, post chatter messages, manage modules, inspect schema — all from the terminal. Built for humans and AI agents alike. Why a CLI? Odoo Toolbox already teaches AI agents the Odoo API through skill modules. But skills require a specific agent setup. The CLI is simpler: install it, and any agent that can run shell commands picks it up immediately via --help. For humans, it’s the fastest way to interact with Odoo without opening a browser. ...

TypeScript library and gateway CLI for the Holded ERP — invoicing, CRM, accounting, projects, and team — designed for AI agents. What It Does Wraps Holded’s API across 9 domains in a single library with an agent-friendly CLI: Contacts: contacts and groups Invoicing: documents (10 types), numbering series, services Accounting: expense/income accounts, chart, daily ledger CRM: funnels, leads, events, bookings Cash: treasuries, payments, remittances Stock: products and warehouses Team: employees, time tracking, clock in/out Taxes: tax configuration Projects: projects, tasks, time tracking Multi-tenant support via profiles — one API key per client, no code changes. ...

Open-source MAPI-to-Gmail bridge for Windows — makes “Send by Email” work with Gmail. What It Does Intercepts MAPI calls (Windows’ “Send by Email” API) and creates Gmail drafts via browser extension. Right-click a file in Explorer → Send → Gmail draft opens with file attached. Every Windows app that can email uses MAPI: PDF printers, accounting software, scanners, CRMs. Gmail has no native MAPI support. The only third-party solution (Affixa) shut down. Go MAPI is the open-source replacement. ...

Agent-agnostic skills that teach AI agents how to use specific tools — structured markdown “man pages” for LLMs. What It Is Skills are standalone markdown files that tell an agent: How to operate a CLI What’s safe to run (safety classification) What patterns to follow What gotchas to avoid CC0-licensed (public domain). Works with pi, Claude Code, Cursor, and any agent that reads markdown. Available Skills Install (8 skills available as of 2026-03-26) ...

Launch commands in a real pseudo-terminal, detach, reattach later. Cross-platform screen / tmux alternative for AI agents. What It Does Creates detached PTY sessions that survive disconnection. Windows (named pipes), macOS, and Linux (Unix sockets). Agents can launch long-running processes, detach, and check back later without blocking. Supports logs --tail/--follow, session replay, Ctrl+A detach in attached mode, and PowerShell compatibility. Quick Start Install (updated 2026-03-26) npx skills add marcfargas/holdpty Install CLI: ...

CLI wrapper around @devcontainers/cli that launches any project in a Dev Container with pi pre-installed and configured. One command to go from project directory to a fully working pi agent session inside a container. What It Does Solves the “works on my machine” problem for AI coding agents. Windows ARM native module failures, path inconsistencies, and tool incompatibilities disappear when pi runs inside a consistent Linux container. Never modifies your project files — merges pi configuration into a temporary devcontainer.json at launch time. Your project’s own devcontainer config (if any) is respected and extended, not replaced. ...