OpenAgent

Next-generation personal and enterprise AI assistant platform powered by LLM, RAG, and agent loops, featuring multi-channel gateways, visual BPMN workflow orchestration, and multi-tenant isolation.

Overview#

OpenAgent is a full-stack AI assistant platform that addresses challenges enterprises and individuals face when adopting LLMs: difficulty in private deployment, lack of knowledge base integration, fragmented multi-platform messaging channels, and the inability to automate complex tasks. Built with a Go monolithic backend + Web frontend architecture, it supports fully self-hosted deployment.

Core Capabilities#

Agent Loops & Tool Calling#

Browser-Use: Drives real browsers for navigation, clicking, form filling, page scraping, and screenshots
Web Search & Fetch: Executes web searches and pulls content directly into the Agent context
Shell Execution: Runs Shell commands and scripts within the Agent loop
Office Automation: Reads and writes Word, Excel, and PowerPoint files
MCP (Model Context Protocol): Connects to any MCP-compatible server via SSE / Stdio / StreamableHTTP to expose tools to the Agent
Transparent Tool Calls: Step-by-step display of tool call parameters and return values

RAG & Knowledge Base#

Document Ingestion: Supports uploading PDF, Word, Excel documents with automatic chunking, embedding, and indexing
Semantic Search: Automatically retrieves the most relevant passages from the knowledge base per conversation
Pluggable Embedding Providers: Supports OpenAI, Azure, Gemini, Qwen, Cohere, Jina, HuggingFace, and local models
Per-Store Isolation: Organizes knowledge into independent Stores, isolated by conversation or application
Vector Database Support: Compatible with Qdrant, Pinecone, Milvus, PgVector, Redis

Model Integration & Scheduling#

Supports 30+ model providers: OpenAI, Azure OpenAI, Claude (Anthropic), Gemini (Google), DeepSeek, Mistral, Grok, Qwen, Doubao, Moonshot, ChatGLM, Baichuan, Ernie, iFlytek, HuggingFace, Cohere, Amazon Bedrock, OpenRouter, and local models

Workflow Automation#

Visual Workflow Builder: BPMN-style visual editor for orchestrating multi-step Agent pipelines
Conditional & Parallel Execution: Supports conditional branching and parallel task execution within workflows
Task Scheduling: Supports scheduled or periodic execution of workflows or Agent tasks

Multi-Channel Messaging Gateway#

Covers 20+ messaging channels: Telegram, Discord, Slack, WhatsApp, Microsoft Teams, WeChat, LINE, Matrix, Signal, Feishu, etc.
Deploy once, run simultaneously across all platforms

Enterprise Features#

Authentication & Isolation: SSO via Casdoor (OIDC / OAuth2 / LDAP / SAML), Multi-tenant isolation by user/organization
Multimedia Processing: Built-in file, image, and video storage management, plus STT and TTS capabilities
Observability: Complete Audit Logs, usage statistics (with interactive charts and heatmaps), activity monitoring, and request logs (full JSON Payload debugging)
REST API + Swagger UI: All features accessible via RESTful API

Architecture Highlights#

Backend entry point is main.go; requests flow through proxy/ layer, dispatched by routers/ to controllers/
chain/ implements Agent loop and chain-of-call logic, combined with tool/ for specific tool execution
mcp/ handles Model Context Protocol integration; embedding/ invokes pluggable embedding models
bpmn/ implements BPMN-spec-based parsing and scheduling engine
Frontend in web/ (Yarn build), with i18n support (i18n/) and CLI entry (internal/cli/)
pkgdocker/ and pkgkubernetes/ provide containerization and cloud-native deployment support

Installation#

Precompiled Binary (Recommended)

macOS / Linux / WSL:

curl -fsSL --proto '=https' --tlsv1.2 \
  https://raw.githubusercontent.com/the-open-agent/openagent/master/scripts/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/the-open-agent/openagent/master/scripts/install.ps1 | iex

Optional env vars: OPENAGENT_VERSION, INSTALL_DIR, BIN_DIR
Access at: http://localhost:14000

Build from Source

go build
cd web && yarn install && yarn start

Docker Deployment Includes Dockerfile, docker-compose.yml, docker-entrypoint.sh for standard containerized deployment.

Use Cases#

Self-hosted private ChatGPT alternative
Enterprise knowledge base management and semantic search
Unified multi-channel intelligent customer service deployment
Office automation (document and browser operations)
Complex business flow visual orchestration
Development assistance and MCP tool integration

Unconfirmed Information#

Default vector database not explicitly specified
Default embedding model not stated in README
Repo Topic tags A2A (Agent-to-Agent) support, but specific implementation is unconfirmed
Topic includes openclaw, not mentioned in README; relationship unconfirmed
Background, members, and funding of the-open-agent organization are not publicly disclosed