S
SkillNav

Percept

AI 与智能体

by getpercept

为 AI agents 提供环境语音智能，通过可穿戴麦克风采集语音，并经由 MCP 触发 agent actions。

5GitHub

什么是 Percept？

为 AI agents 提供环境语音智能，通过可穿戴麦克风采集语音，并经由 MCP 触发 agent actions。

README

Percept — The Context Layer for AI Agents

Give your AI agent ears, eyes, and awareness.

Percept connects your real-world data to your AI agent through hardware capture, tool connectors, and a knowledge graph that enables your agent to take initiative.

What It Does

  • 🎧 Hear — Omi pendant captures ambient audio. Chrome extension captures browser audio. Meetings, conversations, podcasts — all transcribed and searchable.
  • ⌚ Interact — Apple Watch app with push-to-talk, raise-to-speak, and complications.
  • 🔗 Connect — Gmail, GitHub, Linear, Calendar, Slack connectors sync your work context.
  • 🧠 Understand — Knowledge graph connects people, projects, conversations, and events across all sources.
  • ⚡ Act — Initiative engine detects patterns and triggers actions without being asked.

Quick Start

bash
pip install getpercept
percept serve          # Start the audio pipeline
percept sync           # Sync all connected tools
percept status         # Check what your agent knows

Architecture

code
Hardware (Omi, Watch, Chrome)  → Audio Pipeline → Transcription
Tool Connectors (Gmail, GitHub, Linear) → Data Sync
                    ↓
    Knowledge Graph (entities, relationships, temporal)
                    ↓
    Initiative Engine (patterns → actions)
                    ↓
    Your AI Agent (OpenClaw, Claude, GPT, LangChain)

Package Structure

code
percept/
├── percept/                 # Python package
│   ├── audio/               # Audio pipeline (receiver, transcriber, context)
│   ├── core/                # Knowledge graph (graph DB, ingest, query, temporal)
│   ├── connectors/          # SDK + Gmail, GitHub, Linear, Calendar, Slack
│   ├── pipeline/            # Orchestration (sync → KG → signals → initiatives)
│   ├── initiatives/         # Rules engine, pattern matching, actions
│   ├── mesh/                # Team agent shared context
│   ├── memory/              # Entity extraction, FTS5 search
│   ├── mcp/                 # MCP server for Claude Desktop
│   └── cli.py               # Unified CLI entry point
├── src/                     # Legacy audio module (original package root)
├── watch-app/               # Apple Watch app (Swift)
├── extension/               # Chrome extension (browser audio capture)
├── web/                     # Landing page
├── docs/                    # Documentation
└── tests/                   # Test suite

Connectors

ConnectorStatusWhat It Captures
Omi Pendant✅ LiveAmbient audio, meetings, conversations
Apple Watch✅ TestFlightPush-to-talk, raise-to-speak
Chrome Extension✅ BuiltBrowser tab audio (meetings, YouTube, podcasts)
Gmail✅ LiveEmails, threads, contacts
GitHub✅ LivePRs, issues, commits, reviews
Linear✅ LiveTickets, projects, team activity
Calendar🔶 ReadyEvents, attendees (needs Google API enabled)
Slack📋 PlannedMessages, channels, threads

CLI Commands

bash
percept sync                    # Run all connectors → KG → initiatives
percept sync --connector gmail  # Single connector
percept status                  # KG stats + connector health + initiatives
percept query "..."             # Search the knowledge graph
percept serve                   # Start the audio pipeline + API
percept connectors              # List installed connectors
percept initiatives             # List triggered initiatives

Works With Any AI Framework

Percept is framework-agnostic. It provides context via:

  • CLI — percept sync, percept query
  • MCP Server — 8 tools for Claude Desktop integration
  • REST API — HTTP endpoints for any framework
  • Python SDK — from percept.core.graph.database import GraphDB

License

MIT

常见问题

Percept 是什么？

为 AI agents 提供环境语音智能，通过可穿戴麦克风采集语音，并经由 MCP 触发 agent actions。

相关 Skills

Claude接口

by anthropics

Universal
热门

面向接入 Claude API、Anthropic SDK 或 Agent SDK 的开发场景，自动识别项目语言并给出对应示例与默认配置，快速搭建 LLM 应用。

✎ 想把Claude能力接进应用或智能体，用claude-api上手快、兼容Anthropic与Agent SDK，集成路径清晰又省心

AI 与智能体
未扫描114.1k

RAG架构师

by alirezarezvani

Universal
热门

聚焦生产级RAG系统设计与优化，覆盖文档切块、检索链路、索引构建、召回评估等关键环节，适合搭建可扩展、高准确率的知识库问答与检索增强应用。

✎ 面向RAG落地，把知识库、向量检索和生成链路系统串联起来，做架构设计时更清晰，也更少踩坑。

AI 与智能体
未扫描10.2k

计算机视觉

by alirezarezvani

Universal
热门

聚焦目标检测、图像分割与视觉系统落地，覆盖 YOLO、DETR、Mask R-CNN、SAM 等方案，适合定制数据集训练、推理优化及 ONNX/TensorRT 部署。

✎ 把目标检测、图像分割到推理部署串成完整工程链路，主流框架与 YOLO、DETR、SAM 等方案都覆盖，落地视觉 AI 会省心很多。

AI 与智能体
未扫描10.2k

相关 MCP Server

顺序思维

编辑精选

by Anthropic

热门

Sequential Thinking 是让 AI 通过动态思维链解决复杂问题的参考服务器。

✎ 这个服务器展示了如何让 Claude 像人类一样逐步推理，适合开发者学习 MCP 的思维链实现。但注意它只是个参考示例，别指望直接用在生产环境里。

AI 与智能体
83.4k

知识图谱记忆

编辑精选

by Anthropic

热门

Memory 是一个基于本地知识图谱的持久化记忆系统，让 AI 记住长期上下文。

✎ 帮 AI 和智能体补上“记不住”的短板，用本地知识图谱沉淀长期上下文，连续对话更聪明，数据也更可控。

AI 与智能体
83.4k

PraisonAI

编辑精选

by mervinpraison

热门

PraisonAI 是一个支持自反思和多 LLM 的低代码 AI 智能体框架。

✎ 如果你需要快速搭建一个能 24/7 运行的 AI 智能体团队来处理复杂任务（比如自动研究或代码生成），PraisonAI 的低代码设计和多平台集成（如 Telegram）让它上手极快。但作为非官方项目，它的生态成熟度可能不如 LangChain 等主流框架，适合愿意尝鲜的开发者。

AI 与智能体
6.8k

评论