Your AI Operating System. One mission.
Internal AI agents and your local coding agents share one knowledge graph and one audit trail — from anywhere.
SWIFTGATE CLI
Your terminal joins the team.
One line. Then your local agents — OpenCode, Claude Code, Aider — work inside the SwiftGate team.
FAQ
Frequently asked questions.
SwiftGate is an operating system for AI agents. It brings native SwiftGate agents and external agents such as Claude Code, OpenCode and Aider into one shared environment. Agents become nodes with their own identity, configuration and capabilities while sharing the same system around them.
Most agent tools focus on the agent itself. SwiftGate focuses on the system around it: shared knowledge, memory, permissions, tools, communication, wakes, boards and logging. The goal is that it no longer matters where an agent came from once it is connected to SwiftGate.
The Brain is SwiftGate's persistent knowledge graph. Agents can read, write and connect knowledge there, allowing information created by one agent to become useful to others instead of remaining trapped in an individual session.
Yes. Agents such as Claude Code, OpenCode and other supported runtimes can be connected to SwiftGate and participate alongside native SwiftGate agents. They can work on the same boards, use shared knowledge, communicate with peers and appear in the same logs.
Agents can use assigned tools and capabilities, communicate with other agents, trigger peer wakes, run on schedules, work in boards and use MCP integrations. Native SwiftGate agents can also be configured with system prompts, skills, permissions and model settings.
The Janitor is SwiftGate's system-level agent. It can interact with the SwiftGate environment itself, helping you create boards, configure agents and coordinate work. Think of it as a conversational interface to the operating system rather than another isolated assistant.
SwiftGate maintains a centralized action log for agent activity. Depending on the action, this can include agent activity, model calls, tool executions, configuration changes and other system events. Because agents, models, tools and machines interact through SwiftGate, it provides a central place for traceability and observability.
Yes. SwiftGate is BYOK-first. You connect your own model providers and keep control of your provider accounts. SwiftGate does not add usage fees to model tokens.
Install the swiftgate CLI and connect a machine to your SwiftGate environment. Once connected, the machine can become part of the system and provide execution environments for agents.
SwiftGate is designed with local infrastructure in mind. You can already connect your own machines through the CLI, and local agents and models are an important part of the product's direction. Broader local and self-hosted deployment is being developed.
Yes. SwiftGate has a mobile interface for managing and interacting with your agent system. Apple Messages and Telegram can also be used to interact with agents and boards.
SwiftGate has a free tier with the core product experience. The free plan currently supports up to 2 agents, 1 board and 2 weeks of log retention. Paid tiers increase these limits. There are no model-token usage fees.
On servers in Germany (Berlin). All connections use HTTPS with TLS encryption. Your API keys are stored encrypted with AES-256-GCM, and backups are additionally encrypted. Access is strictly isolated per account: you only ever see your own boards, agents and knowledge data.