FairePlace Public MCP Server
Enables AI agents to access French residential rental tools including lease creation and e-signing, rent control checks, rent revision, and document generation.
README
FairePlace — Public MCP Server 🏠🇫🇷
A public, no-auth, remote MCP server that exposes FairePlace's free French residential rental tools to AI agents (ChatGPT, Claude, Perplexity, Cursor…): create & e-sign a lease (bail), check rent control, revise rent (IRL), compute a security deposit or rental yield, get an address risk report, and generate rental documents (rent receipts, inventory, notices, and more).
Zero LLM inference server-side. The server is a thin, deterministic proxy over FairePlace's public API — the calling agent pays for its own inference; FairePlace only runs Rust/Typst logic. Marginal cost ≈ 0.
Endpoint: https://mcp.faireplace.com/mcp · Registry: com.faireplace/public-tools
Quick start
Add the remote server to any MCP client:
| Client | How |
|---|---|
| Claude (Desktop / claude.ai) | Settings → Connectors → Add custom connector → https://mcp.faireplace.com/mcp |
| Cursor | Settings → MCP → Add → Streamable HTTP → https://mcp.faireplace.com/mcp |
| ChatGPT (Custom GPT) | Import the OpenAPI at https://mcp.faireplace.com/openapi.json as an Action (auth: none) |
| MCP Inspector | npx @modelcontextprotocol/inspector@latest → connect to the URL |
No API key, no account. Ask your agent things like "Check if a 1500 €/month rent is legal at 75001 Paris for a 45 m² 2-room" or "Create a furnished lease for a flat in Paris at 850 €".
Tools (21)
Data & compliance (return JSON):
| Tool | Purpose |
|---|---|
check_rent_control |
Rent control verdict for an address (encadrement des loyers) |
get_reference_rent |
Reference rent (€/m²) for a rent-controlled quarter |
simulate_irl / simulate_irl_by_quarters |
Annual rent revision (IRL), explicit or server-resolved INSEE values |
calculate_deposit_restitution |
Security-deposit cap, restitution deadline, penalties |
simulate_rental_yield |
Gross/net yield + per-regime taxation (LMNP, micro-foncier…) |
check_address_risks |
Natural/technological risk report (Géorisques) |
Lease creation (the conversion tool):
| Tool | Purpose |
|---|---|
create_lease_draft |
Create a lease draft → returns a link to finalize & e-sign it (advanced eIDAS, PAdES-B-LT) on FairePlace |
Document generation (return a link that renders the PDF): generate_quittance, generate_etat_des_lieux, generate_conge, generate_cautionnement, generate_avenant, generate_resiliation_amiable, generate_relance_impaye, generate_inventaire_mobilier, generate_notice_information, generate_reparations_locatives, generate_grille_vetuste, generate_charges_recuperables, generate_bordereau_restitution.
How the lease → signature flow works
agent → create_lease_draft(owner, logement, lease)
→ FairePlace stores an anonymous draft, returns draft_token
→ tool returns https://mcp.faireplace.com/r?draft_token=…
user clicks the link
→ GET /r sets the draft_token cookie on .faireplace.com, 302 → app sign-up
→ on login, FairePlace ingests the draft and opens it on the Signature step
The server never hands back a signable document — signature stays behind FairePlace auth. Every link carries utm_source=mcp for attribution.
Architecture
Intentionally a thin adapter — not a domain application. All business logic (rent-control data, IRL indices, legal document rendering) lives in FairePlace's Rust backend; this Worker only maps MCP tool calls to the public /public/tools/* HTTP endpoints and formats results.
- Runtime: Cloudflare Worker (
McpAgent, Durable Object for session state). - Transport: Streamable HTTP on
/mcp. - Guardrail: native per-IP rate limit (100 req / 60 s).
- PDF delivery:
/pdf/<slug>?p=<base64url payload>serves an HTML page that fetches the PDF browser-side (avoids a same-zone Worker→origin block; no storage — the PDF is a local blob).
Endpoints
| Path | Purpose |
|---|---|
/mcp |
MCP server (Streamable HTTP) |
/openapi.json |
Curated OpenAPI 3.1 for GPT Actions |
/.well-known/ai-plugin.json |
Plugin manifest |
/llms.txt |
Passive AI discovery |
/r |
Draft-resume → sets cookie → sign-up |
/pdf/<slug> |
On-demand PDF rendering (browser-side) |
Development
npm install
npm run typecheck
npm run dev # wrangler dev (local)
npm run inspect # MCP Inspector
npm run deploy # wrangler deploy (needs Cloudflare auth)
Config lives in wrangler.jsonc (vars: API_BASE, WEB_BASE, SELF_BASE, DRAFT_REPRISE_URL, APP_SIGNUP_URL).
Discoverability
- Official MCP Registry:
com.faireplace/public-tools - Glama: connector
com.faireplace/public-tools
License
MIT © 2026 FairePlace (Thinkylab SAS). Not legal advice — an aid for French rental workflows.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。