Deployment

Agent One is a single Go binary. It runs anywhere Go runs — your laptop, a VPS, EC2, or a container. Docker is optional, never required.

Self-Hosted (Recommended)

Build and Run

# Build a static binary
go build -o agent-one ./cmd/agent

# Run with config in current directory
./agent-one

# Or specify a config path
./agent-one --config /etc/agent-one/config.yaml

Cross-Platform Builds

Use GoReleaser for reproducible cross-platform builds:

# Local test build
goreleaser release --snapshot --clean

# Full release (creates GitHub release + binaries)
goreleaser release

Supported targets:

OS	Architecture
Linux	amd64, arm64
macOS	amd64, arm64
Windows	amd64

All builds use CGO_ENABLED=0 for fully static binaries with no external dependencies.

Systemd Service

For always-on deployment on Linux:

[Unit]
Description=Agent One
After=network.target

[Service]
Type=simple
User=agentone
WorkingDirectory=/opt/agent-one
ExecStart=/opt/agent-one/agent-one --config /opt/agent-one/config.yaml
Restart=always
RestartSec=5
Environment=LITELLM_API_KEY=your-key
Environment=GITHUB_TOKEN=your-token

[Install]
WantedBy=multi-user.target

sudo systemctl enable agent-one
sudo systemctl start agent-one

Docker

Docker is optional but convenient for running Agent One alongside LiteLLM.

Dockerfile

FROM golang:1.22-alpine AS builder
WORKDIR /app
COPY go.mod go.sum ./
RUN go mod download
COPY . .
RUN CGO_ENABLED=0 go build -o agent-one ./cmd/agent

FROM alpine:3.19
RUN apk add --no-cache ca-certificates
COPY --from=builder /app/agent-one /usr/local/bin/
COPY config.yaml /etc/agent-one/
CMD ["agent-one", "--config", "/etc/agent-one/config.yaml"]

Docker Compose

Run Agent One + LiteLLM together:

version: "3.8"
services:
  litellm:
    image: ghcr.io/berriai/litellm:main-latest
    ports:
      - "4000:4000"
    volumes:
      - ./litellm_config.yaml:/app/config.yaml
    command: ["--config", "/app/config.yaml"]

  agent-one:
    build: .
    depends_on:
      - litellm
    volumes:
      - ./config.yaml:/etc/agent-one/config.yaml
      - ./data:/data
    environment:
      - LITELLM_API_KEY=${LITELLM_API_KEY}
      - GITHUB_TOKEN=${GITHUB_TOKEN}
    ports:
      - "9090:9090"   # Webhooks
      - "4200:4200"   # Web portal

Deployment Modes

The same binary supports three modes, differentiated by config:

Personal (Default)

Single user, single config, single SQLite database. No authentication needed — you own the binary.

agent:
  name: "My Assistant"

Team

One owner + team members. RBAC with 4 roles.

safety:
  rbac: true

Role	Chat	Config	Hands	Approve Plans	Audit	Manage Members
owner	yes	yes	yes	yes	yes	yes
admin	yes	no	yes	yes	yes	no
member	yes	no	no	no	no	no
guest	no	no	no	no	no	no

Multi-Tenant

Multiple isolated tenants on the same binary. Each tenant gets its own config, database, channels, and budget.

tenants:
  - id: "Esteban"
    config: "tenants/Esteban/config.yaml"
    db_path: "tenants/Esteban/agent.db"
    channels: ["cli", "whatsapp"]
  - id: "cronoss"
    config: "tenants/cronoss/config.yaml"
    db_path: "tenants/cronoss/agent.db"
    channels: ["whatsapp-cronoss", "telegram-cronoss"]

No shared state between tenants. Budget tracked independently. Tenant resolution by channel.

Web Portal

The web portal runs on a configurable port (default 4200):

portal:
  enabled: true
  port: 4200

Access it at http://localhost:4200. In production, put it behind a reverse proxy (nginx, Caddy) with TLS.

Performance Targets

Metric	Target
Binary size	< 30 MB
Startup time	< 2 seconds
Idle memory	< 50 MB
Average response latency	< 3 seconds (cheap model)
Go module dependencies	< 15

Health Monitoring

Agent One exposes health endpoints and supports observability via:

• slog structured logging to stderr
• OpenTelemetry tracing on critical paths (agent loop, LLM calls, tool execution)
• Metrics — message count, latency, token usage, cost, queue depth
• doctor command — diagnose configuration and connectivity issues

./agent-one doctor  # Check LiteLLM connection, MCP servers, channel configs