Skip to content
Agent One Docs

Welcome to Agent One

Agent One is a personal and business AI agent runtime written in Go. It receives signals from messaging platforms (WhatsApp, Telegram, ClickUp, CLI), routes them through any LLM via LiteLLM proxy, executes tools via MCP, and responds — all from a single self-hosted binary.

Why Agent One

Section titled “Why Agent One”
  • Single binary — go build produces one executable. No Docker required.
  • Multi-model — route to Kimi, Claude, DeepSeek, Gemini, Ollama, or any LLM via LiteLLM.
  • Multi-channel — WhatsApp, Telegram, Slack, ClickUp, CLI, webhooks.
  • MCP-native — any MCP server becomes a new capability with 4 lines of config.
  • Self-hosted — you own the binary, the data, and the infrastructure.
  • Chat-configurable — change schedules, budgets, and channels by talking to the agent.

Quick Start

Section titled “Quick Start”
Terminal window
# Prerequisites: Go 1.22+, LiteLLM running on localhost:4000
go build -o agent-one ./cmd/agent
./agent-one                      # Run with CLI channel
./agent-one --tui                # Run with terminal UI

Architecture

Section titled “Architecture”
Signals ──> Event Queue ──> LLM Router ──> Tool Execution ──> Response
                                 │
                          ┌──────┴──────┐
                          │  Agent One  │
                          │  (Go binary)│
                          └──────┬──────┘
                                 │ HTTP (OpenAI format)
                          ┌──────┴──────┐
                          │  LiteLLM    │
                          │  (proxy)    │
                          └──────┬──────┘
                           ┌─────┼─────┐
                         Kimi  Claude  DeepSeek  Ollama ...

Signals (Message, Cron, Heartbeat, Webhook, Hook) flow through an event queue to the LLM router. The router picks a model complexity level (cheap, medium, or smart) and sends the request to LiteLLM. Tool calls execute via MCP servers, and responses route back through the originating channel.

What’s Next

Section titled “What’s Next”