AIAny
AI Agent2026
Icon for item

Paperasse

A collection of Markdown 'skills' that convert LLM-based agents into specialists for French administrative workflows — accounting, taxes, invoicing, notary work and audits. Ships agent-agnostic skills, Qonto/Stripe connectors, FEC output and 2026 e-invoicing guidance.

Introduction

French administrative and tax workflows are highly structured but painfully detailed — the sort of tasks that LLM agents can consistently handle if given reliable, codified rules. Paperasse packages domain knowledge (PCG, CGI/BOFiP references, FEC formats, 2026 e-invoicing rules) into reusable Markdown skills so an agent can perform bookkeeping, simulate tax audits, generate compliant invoices and prepare year-end filings with fewer manual steps.

What Sets It Apart
  • Narrow, jurisdictional focus: encodes French accounting taxonomy (≈800 PCG accounts), liasse/FEC formats (18 columns), and legal references so outputs map to real filing requirements — not just generic bookkeeping advice. This reduces common mistakes (wrong account codes, missing mandatory invoice fields) that are costly in France.
  • Agent-agnostic skillset: distributed as Markdown skills that work with multiple agent platforms (Claude Code, Cursor, Mistral Vibe, Codex-style agents). That makes adoption easier across different LLM tools without vendor lock-in.
  • Practical integrations and tooling: includes connectors for Qonto and Stripe to ingest transactions, scripts to generate FEC, PDFs and liasse drafts, and an evals suite that measures skill impact (the repo reports measurable deltas in automated evals).
  • Compliance-aware workflow: skills check company context (SIREN, regime TVA) before acting and surface expiry/last-updated metadata so the agent prompts for online verification when rules may have changed.
Who it's for — and trade-offs

Great fit if you are a French SME, bookkeeper, accounting firm or developer integrating LLM agents to automate repetitive, rules-based administrative tasks and want a reproducible, auditable skillset. It speeds up closing, invoice generation and routine tax calculations. Look elsewhere if your operations are outside French jurisdiction (rules and forms here are very France-specific), if you need audited-grade legal advice (the repo explicitly warns it does not replace licensed professionals), or if you require turnkey SaaS with hosted data compliance features — Paperasse is an open-source skillset that expects some technical integration and governance from the adopter.

Where it fits

Treat Paperasse as an onboarding layer for LLM agents that need deterministic wiring to French tax/accounting rules: use it to codify workflows and reduce human error, then pair with professional review for final filings.

Information

  • Websitegithub.com
  • Authorsromainsimon
  • Published date2026/02/05

Categories

More Items

Hugging Face
AI Model2026

Open-weights LLM fine-tuned for phone-based voice agents that prioritizes low latency and reliable tool/function calling. Based on NVIDIA Nemotron 3 Nano (30B total, 3.5B active), supports very long contexts (262,144 tokens) and recommends temperature=0 with thinking disabled for deployment.

GitHub
AI Agent2026

A curated collection of production-ready Agent Skills that turn tasks—presentation production, image generation, local KB retrieval, article assembly, and web-design—into agent-loadable skill folders. Uses a SKILL.md contract, supports multiple agent runtimes (Claude Code, Cursor, Codex), and offers modular install paths with pinned release zips.

GitHub
AI Agent2026

Provides a ruleset and skills that make AI coding agents prefer the simplest correct implementation: reuse existing code, prefer stdlib/native features, and only write minimal new code. Cuts generated LOC, tokens, cost and time while preserving validation and safety.