Platforms and tools for AI agents.
Build and operate AI agents with governed models, memory, knowledge, and isolated workspaces. Keep everyday work moving with monitoring, terminal, and launcher tools.
Dedicated services for agent orchestration, model routing, durable context, grounded knowledge, and isolated execution. Deploy them in your own environment and connect the services you need.
Build agents, run them anywhere, review their work.
Configure agents with models approved in Agent Models, skills, MCP tools, and sub-agents. Run them from chat, APIs, Slack, Teams, Telegram, webhooks, or schedules. Control access, tool approvals, and spending limits, and review each run's trace.
MCP tools · Tool approvals · Run traces · Spending limits
One gateway for every model your agents call.
Connect OpenAI, Anthropic, Google, xAI, Bedrock, OpenRouter, or self-hosted endpoints with encrypted keys. Approve the models agents may call, route requests in each provider's native format, and track usage and cost by model, user, agent, and service.
Provider routing · Model approval · Usage & cost ledger · Public catalog
Carry useful context across agents and sessions.
Save rules, preferences, decisions, facts, and lessons with their sources. Recall them in Studio, Claude Code, and Codex, and share them by user, team, or organization. Nightly Dreaming merges and promotes memories, and every change keeps a version.
Durable context · Scoped sharing · Nightly Dreaming · Client handoffs
Give agents knowledge they can trace to a source.
Ingest documents and recordings, then produce transcripts and meeting notes. Search text and the knowledge graph with links to source evidence. Track decisions, tasks, and risks as objects, and review AI-extracted knowledge before it is applied.
Source evidence · Meeting notes · Knowledge graph · Ontology objects
Give agent work a place to run.
Run commands and Codex, Claude, or OpenCode sessions in Docker or Kubernetes sandboxes. Workspaces and sessions persist across suspend and resume. Connect GitHub, push branches and pull requests, and approve merges, releases, and deploys after reviewing the diff.
Persistent workspaces · Coding CLI sessions · Workspace profiles · Approved releases
Monitor your systems, keep coding sessions running, and make everyday actions easier.
Monitor your agents, accounts, and infrastructure.
Follow coding-agent activity and plan usage across accounts, with metrics from servers, Kubernetes clusters, DGX Spark hosts, and custom resources. View the same workspace on desktop, web, iPhone, Apple Watch, Android, or an ESP32 display, and pin it to widgets.
Desktop · Web · iOS · Android · ESP32 · Coding agent activity · Plan usage limits · Resource metrics · Widgets
Close the terminal. The shells keep running.
Keep coding agents running on an always-on machine after you close the terminal or disconnect SSH. Reconnect to the same workspace tabs, search retained output, and see which agents are working, waiting for approval, or done.
macOS · Linux · Persistent sessions · Agent resume · Searchable scrollback · Agent status
A launcher you summon with one shortcut.
A global hotkey opens your own pages of buttons. Run commands, open files and URLs, run scripts, chain actions, or launch applications. On macOS, snippets expand typed keywords in any app. Optional Cloud Sync keeps pages and snippets in step across devices.
macOS · Windows · Custom button pages · Script actions · Text expansion · Cloud sync
Backend development and SRE experience guide how we build, ship, and operate agent systems.
- Agent platforms — Models, tools, memory, knowledge, and workspaces with clear responsibilities.
- IaC & GitOps — Infrastructure defined in code, reviewed in Git, and deployed consistently.
- DevOps & SRE — A shorter path from a commit to production, with reliability that is measured.
Explore the website or browse our repositories.
