feedback-mcp
MCP server for Open Feedback, enabling AI assistants to submit, list, get, update status of, and analyze product feedback via a local HTTP API and JSONL storage.
README
Open Feedback
Reusable feedback collection for Hasna-coded apps.
Open Feedback provides a small HTTP API, TypeScript SDK, CLI, MCP server, and local JSONL storage so apps can collect product feedback without standing up a database first. The local project slug is open-feedback; the GitHub repository is hasna/feedback.
Install
bun add @hasna/feedback
For local CLI usage:
bunx @hasna/feedback init
feedback serve --port 8787
feedback-serve --port 8787
HTTP API
Start the API:
feedback serve --host 127.0.0.1 --port 8787
Set FEEDBACK_API_TOKEN to require bearer-token auth for API requests:
FEEDBACK_API_TOKEN="$YOUR_TOKEN" feedback serve
Submit feedback:
curl -X POST http://127.0.0.1:8787/v1/feedback \
-H 'content-type: application/json' \
-d '{
"appId": "my-app",
"message": "The billing screen should show the invoice PDF sooner.",
"kind": "idea",
"tags": ["billing"]
}'
Useful endpoints:
GET /healthPOST /v1/feedbackGET /v1/feedback?appId=my-app&limit=50GET /v1/feedback/:idPATCH /v1/feedback/:idwith{ "status": "triaged" }GET /v1/statsGET /v1/export.jsonl
SDK
import { createFeedbackClient } from "@hasna/feedback";
const feedback = createFeedbackClient({
baseUrl: "http://127.0.0.1:8787",
token: process.env.FEEDBACK_API_TOKEN,
});
await feedback.submit({
appId: "my-app",
message: "Export fails after selecting a date range.",
kind: "bug",
severity: "high",
context: {
route: "/reports",
version: "2026.07.01",
},
});
Browser apps can collect standard route/device context without a UI dependency:
import { collectBrowserFeedbackContext } from "@hasna/feedback/browser";
const context = collectBrowserFeedbackContext({
version: import.meta.env.VITE_APP_VERSION,
environment: import.meta.env.MODE,
});
For in-process server apps, use local storage directly:
import { LocalFeedbackStore } from "@hasna/feedback/storage";
const store = new LocalFeedbackStore();
await store.createFeedback({
appId: "my-app",
message: "Add CSV export.",
});
CLI
feedback init
feedback doctor
feedback submit "Add export history" --app my-app --kind idea --tag reports --route /reports --app-version 1.2.3 --env production
feedback list --app my-app --search export --since 2026-01-01 --limit 20
feedback show <id>
feedback status <id> triaged
feedback stats
feedback export --format jsonl --until 2026-12-31
Use --api-url and --token to target a remote Open Feedback API instead of local JSONL storage.
feedback doctor checks the package version, local data file path, basic storage permissions, token configuration, and whether the expected binaries are on PATH.
Terminal Slash Commands
For terminal or agent slash-command style workflows, wire the command body to feedback submit and pass the current app slug:
# /feedback Add an activity filter to the inbox view
feedback submit "Add an activity filter to the inbox view" --app my-app --kind idea --tag slash-command
# /bug Export fails after picking a date range
feedback submit "Export fails after picking a date range" --app my-app --kind bug --severity high
The slash-command wrapper should provide --api-url and --token when feedback belongs in a shared deployment.
MCP
Run the MCP server:
feedback-mcp
Available tools:
submit_feedbacklist_feedbackget_feedbackupdate_feedback_statusfeedback_statsexport_feedback
Storage
By default, Open Feedback writes JSONL to:
~/.hasna/feedback/feedback.jsonl
Override the directory with FEEDBACK_DATA_DIR.
App Integration
See docs/app-integration.md for browser, server, CLI, and MCP integration examples.
Development
bun install
bun run typecheck
bun test
bun run build
Security
Open Feedback redacts common credential patterns and sensitive metadata keys before storing feedback. Treat feedback exports as potentially sensitive product data. Do not commit feedback JSONL files or API tokens.
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
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