Chat App
The Chat App is the user-facing workspace of SecondStack: a single chat interface that spans every model your organization enables. It combines multi-model chat, image generation, voice input, file and image attachments, retrieval over your own documents, and an agentic mode — without sending data to a third-party SaaS.
Agents
Chat Models
Image Generation
The model picker groups options by capability — agents, chat models, and image generation — across providers.
Multi-model from one interface
Section titled “Multi-model from one interface”Users switch between providers in a single picker instead of juggling separate tools:
- Chat models — Anthropic Claude, OpenAI GPT-5 family, and Google Gemini are available out of the box.
- Image generation — generate and edit images with dedicated image models in the same workspace.
- Local models — self-hosted open-weight models are supported through vLLM when you run your own GPU inference (enabled by configuration).
Administrators group models by capability and enable or restrict them per team, so each group only sees the models it is allowed to use. See API Routing for how access control works.
Beyond plain chat
Section titled “Beyond plain chat”- File attachments and vision — upload documents and images; the platform extracts and reads their content so models can answer over them.
- Voice input — dictate prompts with push-to-talk speech-to-text.
- Knowledge Collections — attach named document sets to any chat for retrieval-augmented answers. See Knowledge & RAG.
- MCP tools — let chats reach into internal systems and APIs through Model Context Protocol tools, scoped to a user, team, or the whole organization.
- Cowork Agent mode — turn any chat into an agentic session that runs Claude Code in a sandboxed container. See the Cowork Agent.
Built for teams
Section titled “Built for teams”- System prompts and personas standardize how the AI behaves for specific workflows, without locking individuals out of free-form use.
- Shareable chat templates let teams reuse proven setups.
- Temporary chats keep sensitive conversations out of history when needed.
- Thread forking and full-text search make long-running work easy to branch and find.
- The Chat App installs as a progressive web app (PWA) for a native-like experience on desktop and mobile.
Related
Section titled “Related”- Cowork Agent — agentic sessions inside the Chat App.
- Knowledge & RAG — ground answers in your own documents.
- Cost Management — budgets and usage tracking for every chat.