Getting Started

Trident is a collaborative workspace for working on projects with multiple AI models. Connect directly to Anthropic, OpenAI, or Google, or use models deployed through Amazon Bedrock, Google Vertex AI, and Azure OpenAI. Each thinks and performs a little differently, while your documents and images stay together in one workspace.

Quick Start

1. Connect a provider: Open Settings > Providers and configure at least one direct API or cloud platform. Secrets are encrypted and stored locally on your device.
2. Create a project: From the home screen, click New Project. Give it a name, an optional description, and optionally choose a workspace directory on disk.
3. Start chatting: Open your project to land in the dual-panel chat view. Choose a model, type a message, and the AI will respond in real time.
4. Build up the project’s assets: Create documents yourself with Cmd+N (or the File > New Document menu), or ask any model to draft one for you. Ask a model for an image and it’ll open a picker so you can choose which image model to use. Everything lands in the same project.

Overview

A Trident project is a shared workspace between you and the models you choose to work with. The main project view runs two chat panels side by side so you can drive two models at once (often from different providers) to compare takes, divide the work, or get a second opinion without switching context.

Each project has two kinds of assets: documents and images. Both you and any AI model can create them, and every model in the project sees the same gallery and can read or edit documents you attach to a conversation, including documents that another model originally wrote. This is the core of collaboration in Trident: one model can draft a chapter, a plan, or a spec, and you can hand it to a different model (often from a different provider) for review, editing, or a second take. Documents are written in Markdown and can be saved automatically with autosave enabled, so attaching one to chat gives the receiving model full context while you keep working.

Your Data

All Trident files (projects, documents, and images) are stored locally on your computer in a Trident folder in your home directory (~/Trident/). Trident does not sync that library to its own cloud service. Chat prompts and attached content are sent to the AI connection you choose, and provider secrets are encrypted on-device. Semantic indexing uses OpenAI separately when enabled.

Supported AI Providers

Trident connects in two ways:

• Direct APIs: Anthropic, OpenAI, and Google Gemini, using your own API key for each. Trident lists whatever chat-capable models your key has access to, so new releases appear as soon as the provider offers them.
• Cloud platforms: Amazon Bedrock, Google Vertex AI, and Azure OpenAI, using the models, inference profiles, and deployments already provisioned in your organization’s account. You choose which of them Trident should offer when you set up the connection.

Trident works out what each model supports (extended thinking, image input, context size) and adapts the chat controls to match.