Managed profiles
Fast, Ultra, Creative, and Agent profiles give each workspace a stable starting point.
YOUR AI WORKSPACE · BETA
More models.
One place to think.
Write, research and work through ideas with OpenAI, Claude and Gemini. Bring your own context. Keep the conversations that move you forward.
Free plan · 300 requests/month · No monthly charge
OpenAIAnthropicGoogle
TR Expert workspaceExample conversation
＋ New chatChatSearchGenerateAgentsCode✧ Companion
YOUR CHATSResearch notesIdeas for a launch
✧ TR CompanionChoose a model ⌄
Help me turn these research notes into a clear plan.
✧ TR Companion
Let’s make the next step clear.
Start with the decision you need to make. Then separate what your notes support from what still needs checking.
01 Define the question02 Review the evidence03 Choose the next action
✧ Your context, saved by you
Ask a follow-up…↑
A LITTLE MORE PERSONAL
An assistant that starts
with your context.
Tell TR Companion what you’re working on and how you like to work. Save only what you choose, and bring selected information from your accounts into the conversation.
Meet your CompanionYOUR SPACE
01
Goals & preferencesContext you can edit or delete.
02
Connected accountsGmail, Drive, Calendar and GitHub.
03
You choose what to shareRead, review, then send to Companion.
MODEL CHOICE, MADE VISIBLE
Choose the route.
Know what answered.
OpenAI
Direct model choice for chat, reasoning, research, and code workflows.
Anthropic
Claude models for writing, analysis, and tool-assisted work.
Gemini models for fast research, long-context work, and image-capable tasks.
Exact model availability depends on plan, provider configuration, and task capability.
FIVE WORK MODES · ONE EVIDENCE TRAIL
Different models.
One accountable context.
Chat
Work through questions and drafts in a persistent conversation.
Search
Research the web and keep sources connected to the answer.
Generate
Turn a prompt into focused written output inside the workspace.
Agents
Run tool-assisted work with visible execution and approval controls.
Code
Use models selected for programming and technical tasks.
One governed workspace
Chat, Search, Generate, Agents, Code, and Compare share the same authenticated working environment.
Comparison stays inspectable
Choose available models deliberately, compare outputs, and inspect the evidence before using a result.
Human approval stays in the loop
Tool-assisted work exposes execution state and keeps consequential actions behind explicit approval controls.
FROM PROMPT TO DECISION
A reviewable path,
not a black box.
Ask once
Start in Chat, Search, Generate, Agents, or Code with the context your team already has.
Compare deliberately
Choose the models available to your plan and inspect their answers side by side.
Check the evidence
Keep web sources and execution traces attached to the output they support.
Approve the action
Consequential tool actions remain visible and wait for explicit human approval.
IMPLEMENTED BOUNDARIES
Private work stays behind authentication.
Clerk handles account authentication. Conversation endpoints verify the signed-in user before reading, changing, or deleting saved work.
Model requests are made through server routes, so provider credentials are not sent to the browser. The model selected in the workspace determines where a prompt is processed.
Supported chat and research can use a hosted fallback. The owned Sovereign Core provides durable run state, policy-controlled tools, approval checkpoints, and isolated execution when those capabilities are enabled and healthy.
BEFORE YOU START
Straight answers.
Can I start without paying?
Yes. The Free plan includes core chat, search, starter models, and up to 300 requests per month.
Which models are available?
The catalog includes managed TraceRemove profiles plus supported OpenAI, Anthropic, and Google models. Availability depends on plan and task capability.
Where is my prompt processed?
Your selected model determines the provider route. Requests leave the browser through authenticated server routes; provider credentials are never sent to the client.
What happens if Sovereign Core is offline?
Supported chat, grounded Search, and prompt-only agent work can use the hosted fallback. Durable memory, isolated execution, files, and approval-controlled tools remain unavailable until the owned runtime is healthy.
READY WHEN THE WORK IS