A Terminus app is a browser application that runs inside a desktop window. Apps are multi-file bundles — HTML, CSS, JavaScript — served on isolated origins at {slug}.apps.terminus.build.

Built-in apps include Notes, Calendar, Maps, and Files. Developers can build and publish their own apps to the Store, where any user can install them.

Apps use the Terminus SDK (@timi/app-sdk) to access managed collections, object storage, user data, notifications, connector calls, and lifecycle hooks — all without running server code.

terminus init my-app --kind appcd my-app && npm installterminus dev       # local runtime with fixture datanpm run dev        # hot reload; vite proxies /_terminusterminus publish   # build, validate, publish to the Store

An agent is an autonomous worker that runs on your account. You define what it does, give it a schedule or a trigger, and it runs 24/7 on Terminus infrastructure.

Each agent has an address ({publisher}/{slug}), a description of its purpose, and a set of tool grants (like gmail.read, web.search, or data.write). Agents are created in the Studio editor or from the CLI.

The Creations list shows all your agents in one place — their addresses, status, and when they last ran. You can develop locally with terminus dev, then publish with terminus publish.

A skill is a reusable capability — a prompt template, a tool, or a pipeline — that agents and apps can discover and invoke. Skills are published to the Terminus catalog and searchable by anyone.

Skills can be created in the Studio editor with a visual test bench, or authored locally and published via the CLI. Each skill gets a public profile at terminus.build/skills/{publisher}/{slug}.

# Search the catalogcurl -s "$TERMINUS_API/v1/public/search?q=pdf+extraction&limit=10" \  -H "authorization: Bearer $TOKEN" # Run a skill by addresscurl -s "$TERMINUS_API/v1/skills/by-address/acme/summarize/run" \  -H "content-type: application/json" \  -H "authorization: Bearer $TOKEN" \  -d '{"input":{"text":"..."}}'

Connectors link Terminus to your existing tools — Gmail, Slack, GitHub, Linear, and others. Once connected, agents and apps can call these services through the Terminus gateway with proper authorization.

Connector setup uses standard OAuth (PKCE). The user grants access once; Terminus manages token refresh and scoping. Connection credentials are never exposed to app JavaScript — all calls are brokered server-side.

Developers can register external services with an OpenAPI spec. Terminus handles discovery, permissions, credential brokering, and metering.

The Terminus gateway exposes an OpenAI-compatible chat completions endpoint that routes to your selected model provider. You choose the model; Terminus handles routing, metering, and billing.

# List available modelscurl -s "$TERMINUS_API/v1/terminus/models" \  -H "authorization: Bearer $TOKEN" # Chat completioncurl -s "$TERMINUS_API/v1/terminus/chat/completions" \  -H "content-type: application/json" \  -H "authorization: Bearer $TOKEN" \  -d '{    "model": "openai:gpt-4.1",    "messages": [{"role": "user", "content": "hello"}]  }'

The gateway supports streaming, tool use, and multi-turn conversations. Usage is tracked per-request and visible in the Billing section of your account.

The terminus CLI is the developer's entry point. It handles authentication, local development, publishing, and diagnostics.

terminus login          # browser-approved device sessionterminus status         # show current session and accountterminus init <name>    # scaffold an app, agent, or skillterminus dev            # local runtime with hot reloadterminus publish        # build and publish to the Storeterminus logs <name>    # tail production logsterminus doctor <name>  # diagnose common issuesterminus logout         # revoke this device's session

Device sessions expire after seven days and can be revoked from Account → Authenticated devices. The CLI never stores a reusable API key.

Terminus uses email-based signup with a six-digit verification code, followed by a password. Google OAuth is also available as a login method.

Interactive clients (the browser, CLI, and Desktop app) use device-bound sessions. The CLI runs terminus login, you approve in the browser, and the session is stored locally.

Sessions are short-lived and revocable. There are no long-lived API keys for user accounts. Agent runtimes receive short-lived, capability-scoped child tokens from their parent device session.

Your data lives in your own Terminus data directory — per-user, per-app directories with explicit consent grants. Apps store state as files in your space, not in app-owned database rows.

This means you can replace an app's interface without losing your data, export everything at any time, and see exactly what each app has stored.

The Terminus Store is the marketplace for apps, skills, agents, and services. Every published artifact gets a public profile page with description, screenshots, and usage information.

Publishers are identified by their username. The Store supports search, categories, and community features like likes and comments.

Publishing is open to all users via the CLI or Studio editor. Each published artifact receives a stable address at terminus.build.

For questions, integration help, or to report an issue, reach us at hello@terminus.build.