tillys-omi-mcp

MCP音视频

提供方 tilly-gdf

音视频 stdio · 本地进程 API Key 社区 / 第三方 Claude DesktopClaude
查看接入方式

A custom MCP server that connects Claude and other AI assistants to the Omi platform via stdio transport and dual authentication, offering 38 tools for managing memories, conversations, action items, goals, transcript editing, and speaker management.

一句话选型结论

按资料字段整理

tillys-omi-mcp 属于音视频类 MCP 服务器,stdio · 本地进程,API Key,可直接接入 Claude Desktop、Claude,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类音视频 传输方式stdio · 本地进程 鉴权API Key 部署方式本地部署
GitHub 星标 3 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

音视频
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

Local Deployment MCP Server Omi Platform JavaScript Memory Management Conversation Management Action Item Management Goal Tracking Transcript Editing Firebase Auth

支持客户端

适用场景

Local DeploymentMCP ServerOmi Platform

安装与配置

stdio · 本地进程 API Key
客户端在本机拉起子进程,数据不出本机
node /path/to/omi-mcp-server.js

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 node,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
转写与字幕

把音视频转成文字和字幕。

音视频处理

裁剪、转码、拼接素材。

内容分发

把成品发到对应平台。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Claude,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 使用前要准备好API Key对应的凭据,并按最小权限配置。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在音视频场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Claude 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 不方便管理密钥或第三方授权的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
tillys-omi-mcp当前 未标注 stdio · 本地进程 本地部署 有 2026-02-24
Ableton MCP B7b69a6a 16 stdio · 本地进程 本地部署 有 2026-05-17
Mini Max MCP 8 SSE · 服务端推送 远程部署 有 2026-05-18
LocalAI 未标注 stdio · 本地进程 本地部署 有 2026-04-11
Youtube E0eb3510 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Voicemode MCP 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

音视频
字段来自项目仓库与公开索引,中文说明由本站标注
提供方tilly-gdf上游标注的维护方 tilly-gdf
分类音视频本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式API Key远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · node本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-02-24上游最后核验时间

详细介绍

8.9 KB
来自项目仓库与上游目录的原文,未做删改

Omi MCP Server

A custom MCP (Model Context Protocol) server that connects Claude to the full Omi platform - 38 tools covering memories, conversations, action items, goals, transcript editing, speaker management, and more.

Why This Exists

Omi provides an official MCP server using SSE (Server-Sent Events) transport. Two problems:

  1. Firewall issues on managed devices. SSE maintains persistent long-lived HTTP connections. On Windows, this triggers "Allow Node.js through Windows Firewall?" prompts. On a work-managed device, that's a non-starter.

  2. No Internal API access. The official MCP only uses the Developer API key. It can't edit transcript text, assign speakers to segments, query conversations by status, reprocess conversations, or manage apps/templates - all of which require Firebase authentication.

This server solves both:
- stdio transport - standard outbound HTTPS requests only, no persistent connections, no firewall prompts
- Dual authentication - Developer API key for standard CRUD + Firebase tokens for Internal API access, on a single server

Prerequisites

  • Node.js 24+ (uses built-in fetch() - zero npm dependencies)
  • Omi account with the Omi mobile app installed
  • Claude Code or Claude Desktop (any MCP-compatible client works)

Setup

1. Get Your Credentials

You need three values:

Developer API Key

  1. Open the Omi mobile app
  2. Go to Settings > Developer
  3. Tap "Create Key"
  4. Copy the key (shown once, starts with omi_dev_)

Keys are case-sensitive - it's omi_dev_ lowercase, not Omi_dev_.

Firebase Refresh Token

This gives access to Omi's Internal API (transcript editing, speaker assignment, conversation lifecycle). If you only need the Developer API tools, you can skip this.

  1. Open omi.me in Chrome and log in
  2. Open DevTools (F12)
  3. Go to the Application tab
  4. In the left sidebar, expand IndexedDB
  5. Find firebaseLocalStorageDb
  6. Click into the object store
  7. Find the entry with your user data
  8. Navigate to: value > stsTokenManager > refreshToken
  9. Copy the refresh token string

The refresh token doesn't expire unless you change your password or explicitly revoke sessions. You only need to do this once.

Firebase API Key

This identifies Omi's Firebase project. It's not a secret (it's embedded in the web app), but you need the correct one to avoid PROJECT_NUMBER_MISMATCH errors.

Important: Omi has two Firebase projects. The key in the GitHub repo (AIzaSyC1U6S-...) is the dev project. Production users need the production key. Check the Omi web app's network requests or source to find the current production key.

2. Register the Server

Claude Code (CLI)```bash

claude mcp add --scope user omi \
node /path/to/omi-mcp-server.js \
-e OMI_DEV_API_KEY=omi_dev_your_key \
-e OMI_FIREBASE_API_KEY=your_firebase_api_key \
-e OMI_FIREBASE_REFRESH_TOKEN=your_refresh_token


Replace `/path/to/` with the actual path to `omi-mcp-server.js`.

#### Claude Desktop

Add to your config file (`%APPDATA%\Claude\claude_desktop_config.json` on Windows, `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS):```json
{
  "mcpServers": {
    "omi": {
      "command": "node",
      "args": ["/path/to/omi-mcp-server.js"],
      "env": {
        "OMI_DEV_API_KEY": "omi_dev_your_key",
        "OMI_FIREBASE_API_KEY": "your_firebase_api_key",
        "OMI_FIREBASE_REFRESH_TOKEN": "your_refresh_token"
      }
    }
  }
}

3. Verify

After registration, your MCP client should list all 38 tools. Try a simple read operation:Get my latest 5 Omi memories

Tool Reference

Developer API Tools (API Key Auth)

These tools use your OMI_DEV_API_KEY for authentication.

Memories (5 tools)

Tool Description
get_memories List memories with pagination and category filter
create_memory Create a single memory (fact, insight, note)
create_memories_batch Create up to 25 memories at once
update_memory Update content, visibility, tags, or category
delete_memory Permanently delete a memory

Conversations (6 tools)

Tool Description
get_conversations List completed conversations with filters
get_conversation Get a single conversation by ID (with optional transcript)
create_conversation Create from plain text (runs full processing pipeline)
create_conversation_from_segments Create from structured transcript segments with speaker/timing data
update_conversation Update title or discard status
delete_conversation Permanently delete a conversation

Action Items (5 tools)

Tool Description
get_action_items List action items with status/date/conversation filters
create_action_item Create a standalone action item
create_action_items_batch Create up to 50 action items at once
update_action_item Update description, completion status, or due date
delete_action_item Permanently delete an action item

Goals (6 tools)

Tool Description
get_goals List active goals (boolean, scale, or numeric types)
create_goal Create a new goal (max 3 active)
update_goal Update title, values, or unit
update_goal_progress Quick progress value update
get_goal_history Get progress history over time (up to 365 days)
delete_goal Delete a goal

Internal API Tools (Firebase Auth)

These tools use Firebase ID tokens. They require OMI_FIREBASE_API_KEY and OMI_FIREBASE_REFRESH_TOKEN.

Conversation Lifecycle (3 tools)

Tool Description
get_conversations_by_status List conversations by status: in_progress, processing, merging, completed, failed
process_in_progress_conversation Force-process a stuck in_progress conversation
reprocess_conversation Regenerate title/overview/action items from transcript

Search & Merge (2 tools)

Tool Description
search_conversations Full-text search across all conversations
merge_conversations Combine fragmented recordings into one conversation

Transcript Editing (2 tools)

Tool Description
edit_segment_text Fix transcription errors in a specific segment
assign_segments Bulk assign speaker identity (is_user or person_id)

App/Template Management (5 tools)

Tool Description
get_enabled_apps List installed apps/templates
get_app_details Get full app details by ID
enable_app Install an app/template
disable_app Disable without deleting
delete_app Permanently delete (user-owned only)

People/Speaker Profiles (4 tools)

Tool Description
get_people List all speaker profiles with speech samples
create_person Create a new speaker profile
update_person_name Rename a speaker profile
delete_person Delete profile and all speech samples

Troubleshooting

"Allow Node.js through Windows Firewall?" prompt

You're using the SSE-based official server, not this one. This stdio server makes standard outbound HTTPS requests only - no firewall prompts.

PROJECT_NUMBER_MISMATCH error

You're using the wrong Firebase API key. Omi has two Firebase projects (dev and production). Make sure your OMI_FIREBASE_API_KEY matches the Firebase project your account is on. If your account is on production omi.me, you need the production key - not the dev key from the GitHub repo.

API key not working / 401 errors

  • Dev API keys are case-sensitive: omi_dev_ (lowercase), not Omi_dev_
  • Check that the key is correctly set in your environment variables
  • Verify the key hasn't been revoked in the Omi mobile app

assign_segments returns 404

The endpoint URL format matters. This server uses the Internal API path (/v1/conversations/{id}/segments/assign), not the Developer API path (/v1/dev/user/conversations/{id}/segments/assign). If you've modified the code, verify the path.

Firebase token refresh fails

Your refresh token may have been revoked (password change, session revocation). Re-extract it from omi.me browser IndexedDB following the setup steps above.

Conversations stuck in "processing" status

Use get_conversations_by_status to find them, then process_in_progress_conversation or reprocess_conversation to push them through. This commonly happens after Bluetooth disconnects or app crashes.

Environment Variables

Variable Required Description
OMI_DEV_API_KEY Yes Omi Developer API key (omi_dev_xxx)
OMI_FIREBASE_API_KEY For Internal API tools Firebase project API key
OMI_FIREBASE_REFRESH_TOKEN For Internal API tools Firebase refresh token from browser
OMI_API_BASE_URL No Override API base URL (default: https://api.omi.me)

License

MIT - see LICENSE.

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
tillys-omi-mcp 怎么安装?

在客户端按下面的安装命令接入:node /path/to/omi-mcp-server.js。装完在客户端配置里引用同一个包或端点。

tillys-omi-mcp 怎么在 Claude Desktop、Claude 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

tillys-omi-mcp 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

tillys-omi-mcp 需要 API Key 吗?

资料标注需要API Key认证,使用前准备好对应凭据,并按最小权限配置。

tillys-omi-mcp 支持哪些客户端?

资料标注支持 Claude Desktop、Claude;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

tillys-omi-mcp 能提供哪些能力?

资料显示能力标签包括 Local Deployment、MCP Server、Omi Platform、JavaScript、Memory Management、Conversation Management。具体工具清单以官方文档为准。

tillys-omi-mcp 需要什么运行环境?

运行时为 node;属于本地进程,需要在本机装好运行环境。

tillys-omi-mcp 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-02-24。