# Generalbots > Multi-agent AI platform for autonomous agents, agentic AI orchestration and LLM workflow automation. 80 self-hosted apps - Chat, CRM, Mail, Drive, Calendar, Advanced RAG. Open source, built in Rust. - **Type:** MCP server - **Install:** `agentstack add mcp-generalbots-generalbots` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [generalbots](https://agentstack.voostack.com/s/generalbots) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [generalbots](https://github.com/generalbots) - **Source:** https://github.com/generalbots/generalbots - **Website:** https://generalbots.org ## Install ```sh agentstack add mcp-generalbots-generalbots ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # General Bots **Multi-agent AI platform for autonomous agents, agentic AI orchestration and LLM workflow automation.** Self-hosted, open source, and built in Rust. General Bots gives you 80 sovereign applications — Chat, CRM, Mail, Drive, Calendar, Meetings, Documents and Advanced RAG — that run as one platform on your own infrastructure. Your cloud, your data, your rules. [Website](https://generalbots.org) · [Documentation](https://docs.generalbots.org) · [Features](https://generalbots.org/features/) · [Blog](https://generalbots.org/blog/) · [BotBook](./botbook) [](./LICENSE) [](https://www.rust-lang.org/) [](https://github.com/generalbots/generalbots/releases) [](https://github.com/generalbots/generalbots/graphs/contributors) [](https://github.com/generalbots/generalbots/stargazers) [](https://github.com/generalbots/generalbots/network/members) [](https://github.com/generalbots/generalbots/issues) [](https://github.com/generalbots/generalbots/commits/main) [](https://github.com/generalbots/generalbots/pulls) [](#getting-started) [](#why-general-bots) [](#why-general-bots) [](#why-general-bots) [](#license) --- ## What it looks like Two surfaces on one platform: the desktop suite where people work, and the cloud console where an organisation is run. ### Desktop suite — `:3000` Every application opens as a window **inside one desktop shell**, so Chat, CRM, Drive, Mail, Calendar and Meetings sit side by side instead of in separate browser tabs. A theme manager switches the whole workspace between light, dark and wallpapered variants; the capture above is the light one. The suite is a plain web client served by `botui`, and the same client runs unchanged inside the Tauri wrapper (`botapp`) on desktop and mobile. ### Cloud console — `:4000` The control plane for running General Bots as a service: the **store**, **plans** and **offers**, organisation and workspace management, and the usage dashboard. It shares the brand's dark-first visual system, and the capture above is that theme. The console itself never handles credentials — login and signup live on their own surface at `:5000`. --- ## Why General Bots Most "AI agent" tooling stops at a chat window. General Bots is the platform underneath it: an orchestration layer that connects frontier or local LLMs to real business systems, real documents and real workflows. - **Multi-agent, not multi-chat.** Orchestrate several bots with distinct personalities, tools and knowledge bases from one dashboard. - **Agentic AI workflows.** Autonomous task execution that plans, calls tools and reports back — in chat or over WhatsApp. - **Sovereign by default.** Runs entirely offline or air-gapped. No data leaves your network unless you route it out. - **Model-agnostic.** DeepSeek, Qwen, GLM, Kimi, MiniMax, Yi, Step, Doubao, OpenAI, Anthropic — or any OpenAI-compatible endpoint, including a local llama.cpp server. - **Advanced RAG.** Agentic RAG, Graph RAG, persistent memory and multi-modal retrieval over PDFs, Word, Excel and web pages. - **Low-code by design.** Bot behaviour is defined in BASIC dialogs and documents, so domain experts can contribute without a computer science degree. You write this, and the bot is live: ```basic ' start.bas USE KB "product-manual" USE TOOL "create_ticket" ADD_SUGGESTION "Check my order" ADD_SUGGESTION "Open a support ticket" TALK "Hi! I can check orders or open a ticket for you." ``` --- ## Applications Every application is a self-contained module that can be deployed alone or as part of the whole platform. This repository ships **80 suite applications** and **113 backend crates**; the catalogue below is the platform map, and every entry has its own icon in [`.github/svg/`](./.github/svg). | | Application | What it does | |---|-------------|--------------| | | **Server**Core Platform | Web server, WebSocket messaging, REST API and request routing. Serves both the API and the web interface. | | | **Auth / Identity**Core Platform | OAuth 2.0, OpenID Connect, JWT validation and session management, with role-based access control on every endpoint. SSO-ready. | | | **Shared / Database**Core Platform | PostgreSQL with managed schema definitions, shared models and common utilities used across every application. | | | **AI Engine**Core Platform | LLM provider orchestration across DeepSeek, Qwen, GLM, Kimi, MiniMax, Yi or any OpenAI-compatible API. Automatic failover, token counting, streaming and cost tracking. | | | **Drive / Storage**Core Platform | S3-compatible object storage with upload, download, versioning and automatic indexing for search and RAG. Works with MinIO, AWS S3 and Wasabi. | | | **Dashboards / Analytics**Core Platform | Real-time dashboards covering system health, business KPIs and custom visualisations. Exportable and embeddable. | | | **AI Search**Capabilities | Retrieval-augmented generation over PDFs, Word and Excel with sub-second semantic retrieval and cited answers. | | | **Bot Factory**Capabilities | Rapid prototyping and multi-bot orchestration. Run many bots with distinct personalities from one dashboard. | | | **Broadcast**Capabilities | Omnichannel outbound messaging with AI-driven personalisation across WhatsApp, Telegram and SMS. | | | **Content Generation**Capabilities | Generate on-brand, SEO-optimised content in a consistent brand voice. | | | **APIs in BASIC**Capabilities | Build REST endpoints using simplified BASIC syntax, for legacy integration and rapid delivery. | | | **LLM Tools**Capabilities | Give the model real capabilities: web search, calculation and custom API calls. | | | **Talk to Data**Capabilities | Query SQL databases, Excel files and CSVs in plain language. | | | **Training**Capabilities | Ingest institutional knowledge from Word, Excel and PDF documents with no coding. | | | **Advanced RAG**Capabilities | Agentic RAG, Graph RAG, persistent memory and multi-modal retrieval over the knowledge base. | | | **Calendar**Business & Productivity | CalDAV integration with event creation, conflict detection and automated reminder workflows. Syncs with Google, Outlook and Apple Calendar. | | | **Email**Business & Productivity | IMAP/SMTP integration with automatic templating, attachment handling and trigger-based workflows across inbox, starred, sent and scheduled folders. | | | **Meet / Video**Business & Productivity | Video conferencing with screen sharing, recording, transcription and auto-generated meeting notes. | | | **Documents, Sheets & Slides**Business & Productivity | Office-compatible document processing. Read and generate Word, Excel and PowerPoint files from templates. | | | **WhatsApp & Teams**Business & Productivity | WhatsApp Business API and the MS Teams bot framework, with template messages, proactive notifications and rich media. | | | **People / CRM**Business & Productivity | Contact management with CRM-style relationship tracking, tagging, deal pipelines and automated follow-up reminders. | | | **Knowledge Base**AI & Intelligence | Automatic document ingestion with chunking, embedding and semantic search. PDFs, Word files and web pages all become queryable. | | | **Web Automation**AI & Intelligence | Headless browser automation. Scrape sites, fill forms, capture screenshots and trigger workflows when a page changes. | | | **Search**AI & Intelligence | Full-text and semantic search across everything indexed, combining keyword and vector matching with faceted filtering. | | | **Security**Operations | Encryption, threat detection, audit logging and compliance reporting, built to LGPD, GDPR and HIPAA expectations. | | | **Monitoring**Operations | System health monitoring for CPU, memory, disk and network, with alert thresholds, escalation and uptime tracking. | | | **Analytics**Operations | Event tracking, funnel analysis and conversion metrics across every channel. | | | **RBAC & Multi-Tenancy**Enterprise | Fine-grained role-based access control and multi-tenant workspaces with complete data isolation between organisations. | | | **On-Premise Deployment**Enterprise | Runs entirely inside your own infrastructure with no cloud dependency and no data leaving the network. Air-gapped deployments supported. | | | **Compliance (LGPD, GDPR)**Enterprise | Data subject requests, right to erasure, audit trails, retention policies and anonymisation tooling. | Beyond the in-browser suite, the same agent layer drives **WhatsApp Business**, **MS Teams**, **Telegram** and **email**. --- ## Architecture Two Rust services, one workspace, thirteen platform services behind them. The full picture, top to bottom. **Clients and channels** reach **Caddy**, which routes to **botui** (three ports, one binary) and **botserver** (`8080`, HTTP and WebSocket). Inside botserver a message runs `channel_entry → consent_gate → start.bas → kb/RAG → api_catalog → tool_exec → sink`, over the BASIC (Rhai) engine, the drive compiler and the app registry that gives the model a UI automation surface. Below that sit the **AI layer** — `botllm` routing across Claude, Bedrock, Vertex, GLM, Kimi, Kiro or a local llama.cpp — the **data services** (PostgreSQL, Vault, Valkey, MinIO, Zitadel, Qdrant, Stalwart, LiveKit) and the **drive tenancy model** (`{org}.gborg → {workspace}.gbai → {bot}`) that `drive_monitors` discovers from object storage at runtime. **botserver** is the platform core: 113 domain crates, one per business domain (`botcrm`, `botcalendar`, `botdrive`, `botlearn`, `botcloud`, …). Business logic belongs there, not in the API surface. **botui** serves the front end on three ports from a single binary: the desktop suite on `3000`, the cloud/SaaS pages on `4000`, and login/signup on `5000`. A message flows as: WebSocket → session → `start.bas` (once per session) → knowledge-base injection → LLM or direct tool execution → streamed response. Message type `6` skips the model entirely and runs the compiled `.ast` through Rhai. The long-form version — every module, the two security crates, the RAG modes — is in **[BotBook → Architecture](botbook/src/02-architecture-packages/architecture.md)**. ### Ports | Port | Service | What it serves | |------|---------|----------------| | 3000 | botui (suite) | Desktop suite and the 80 applications | | 4000 | botui (cloud) | Store, plans, dashboard, organizations | | 5000 | botui (login) | Login and signup — the only auth surface | | 8080 | botserver | REST API and WebSocket | --- ## Getting Started ### Requirements - **Rust** — a recent stable toolchain from [rustup.rs](https://rustup.rs) (edition 2021) - **Git** - **mold** (optional, faster linking) — `sudo apt-get install mold` ### Run it ```bash git clone https://github.com/generalbots/generalbots cd generalbots ./restart.sh ``` `restart.sh` stops anything already running, builds botserver and botui in order, and starts both. BotServer then provisions its own stack — PostgreSQL, Vault, MinIO, Valkey, Zitadel, Qdrant and a local llama.cpp server — and reads every credential from Vault. No global database or Redis install is needed or wanted. Once it is up: - **Suite** → http://localhost:3000 - **Cloud** → http://localhost:4000 - **Login** → http://localhost:5000/login - **API** → http://localhost:8080/health First boot downloads the stack and models, so give it a few minutes. Logs go to `botserver.log` and `botui.log`: ```bash tail -f botserver.log botui.log ``` ### Run just the API ```bash cd botserver cargo run -- --noconsole ``` --- ## Building Bots Bots are directories, not code. A bot lives in Drive as `{bot}.gbai/` and is picked up automatically: ``` mybot.gbai/ ├── mybot.gbdialog/ # BASIC scripts: start.bas, tables.bas, {tool}.bas ├── mybot.gbkb/ # documents to index for retrieval ├── mybot.gbot/ # bot configuration └── mybot.gbdrive/ # generated files and reports ``` Four keywords cover most of it: ```basic USE KB "manual" ' index documents and retrieve from them USE TOOL "create_ticket" ' expose a tool to the LLM TALK "How can I help?" ' reply to the user CALL "other_script" ' run another dialog ``` See the **[BotBook](./botbook)** for the full keyword reference and worked examples. --- ## Repository Layout | Directory | What it is | |-----------|------------| | `botserver/` | Platform core — API, agents, BASIC engine, 113 crates | | `botui/` | Web front end — suite, cloud and login servers | | `botapp/` | Tauri desktop wrapper | | `botlib/` | Shared types and utilities | | `botbook/` | Documentation source (mdBook) | | `bottemplates/` | Ready-made bot templates | | `bottest/` | Cross-crate integration tests | | `botdevice/` | Device and IoT integrations | | `botplugin/` | Browser extension | | `botmodels/` | Data-model tooling | This is a single repository — there are no git submodules. Push to `origin` for the public mirror; pushing to `alm` triggers the CI/CD pipeline that builds and deploys, so confirm first. --- ## Contributing Read **[AGENTS.md](./AGENTS.md)** before opening a pull request. It carries the coding rules, security directives and testing workflow the project holds to: no `unwrap()`/`expect()` in production paths, no `#[allow()]` suppressions, no CDN assets, and files kept under 450 lines. Before adding a `.md` file, search `botbook/` for existing documentation. --- ## License MIT License — see [LICENSE](./LICENSE). Each subproject carries its own copy (`botserver/LICENSE`, `botui/LICENSE`, `botlib/LICENSE.txt`, …). General Bots has been developed as open source since 2017. Built in Rust · 536,000+ lines · generalbots.org ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [generalbots](https://github.com/generalbots) - **Source:** [generalbots/generalbots](https://github.com/generalbots/generalbots) - **License:** MIT - **Homepage:** https://generalbots.org Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** yes - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-generalbots-generalbots - Seller: https://agentstack.voostack.com/s/generalbots - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.