accelo-mcp
MCP server exposing Accelo quote tools (list, get, create, update) with per-user OAuth authentication for natural language interaction.
README
accelo-mcp
A Model Context Protocol (MCP) server for Accelo. It exposes Accelo quote tools to MCP clients (such as LibreChat agents) and authenticates per user via Accelo OAuth, so every action respects that user's own Accelo permissions.
Status: early. Tools:
list_quotes,get_quote,get_deal,create_quote,update_quote.
How auth works
This server is an OAuth proxy:
- To the MCP client it acts as an OAuth Authorization Server (with discovery + dynamic client registration + PKCE).
- To Accelo it acts as an OAuth client (authorization-code grant).
The MCP client never sees the Accelo token. When a user authorizes, they are bounced to Accelo's consent screen; the resulting Accelo access/refresh tokens are stored locally (SQLite) keyed to an opaque subject, and the MCP client is handed one of our opaque tokens that maps back to it. Accelo tokens are refreshed automatically.
Quote fields
create_quote and update_quote model every client-editable Accelo quote
field explicitly, each mapping 1:1 to an Accelo quote key. Pick the field that
matches the user's intent — do not route client-facing body text into
notes.
| Field | Client-facing? | Notes |
|---|---|---|
title |
— | Quote title. Required on create. |
affiliation_id |
— | Company/contact affiliation the quote is for. |
manager_id |
— | Accelo staff member who owns/manages the quote. |
date_expiry |
— | Expiry date as a Unix timestamp (seconds). |
notes |
No (INTERNAL) | Internal notes only. Never put introduction/conclusion/terms text here. |
introduction |
Yes | Introduction section shown at the top of the quote. |
conclusion |
Yes | Conclusion section shown at the bottom of the quote. |
terms_and_conditions |
Yes | Terms and conditions section. |
client_portal_access |
Yes | Client portal access setting. |
History: earlier versions only surfaced
titleandnotesas named params; all other keys were reachable only through an unconstrained freeformfieldsmap. A request to set the conclusion therefore got silently written tonotes(Accelo accepts it, so no error was raised). These fields are now modeled explicitly and the freeform passthrough was removed. Seesrc/mcp.js(editableQuoteFields).
Prerequisites
- An Accelo deployment (e.g.
https://YOURDEPLOY.api.accelo.com). - A registered Accelo Web Application (Accelo: Configurations -> API ->
Register Application). Set its redirect URI to:
https://<your-public-host>/oauth/callback - Node 20+ or Docker.
Configuration
Copy .env.example to .env and fill in:
| Var | Description |
|---|---|
ACCELO_CLIENT_ID / ACCELO_CLIENT_SECRET |
From your Accelo Web App |
ACCELO_BASE_URL |
https://YOURDEPLOY.api.accelo.com/api/v0 |
ACCELO_OAUTH_URL |
https://YOURDEPLOY.api.accelo.com/oauth2/v0 |
PUBLIC_BASE_URL |
Public HTTPS URL of this server |
OAUTH_REDIRECT_URI |
<PUBLIC_BASE_URL>/oauth/callback (must match Accelo) |
OAUTH_SCOPE |
Accelo scopes, e.g. read(all) write(all) |
TOKEN_DB_PATH |
SQLite path (default /app/data/tokens.db) |
PORT |
Listen port (default 8787) |
Never commit .env (it is gitignored).
Run
Docker (recommended)
docker compose up -d --build
curl -s http://127.0.0.1:8787/healthz
The compose file publishes only on loopback (127.0.0.1:8787). Put a TLS
reverse proxy (nginx, Caddy, etc.) in front of it for PUBLIC_BASE_URL.
Node
npm install
npm start
Endpoints
| Path | Purpose |
|---|---|
POST /mcp |
MCP streamable-HTTP endpoint (Bearer auth) |
GET /.well-known/oauth-authorization-server |
AS metadata |
GET /.well-known/oauth-protected-resource |
Resource metadata |
POST /register |
Dynamic client registration |
GET /authorize |
Start auth (redirects to Accelo) |
GET /oauth/callback |
Accelo redirect target |
POST /token |
Token + refresh |
GET /healthz |
Health check |
Using with LibreChat
Add an MCP server pointing at https://<your-public-host>/mcp with OAuth
enabled. LibreChat will discover the OAuth endpoints, register a client, and
prompt each user to authorize Accelo on first use.
License
MIT
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。