Your AI assistant. Your model workflow.
Connect Claude Code, Cursor, Codex, or another compatible MCP client to Tensorant. Ask your assistant to inspect data, prepare a dataset, compare evaluations, or manage a training experiment.
Connect your MCP client, sign in, and choose the access level to grant.
Interactive illustration · sample data
From an instruction to a prepared dataset. See an illustrative assistant workflow, with visible tool calls and scoped access.
Delegate work with clear boundaries.
An assistant acts as you, within one organization. Its access follows your role, the grant you choose, and your plan’s limits.
Connect the assistant you use
Use the remote MCP endpoint with supported clients, including Claude Code, Cursor, and Codex. Follow the client-specific setup in the documentation.
Explore the featureChoose Read, Build, or Full access
Start with inspection-only access. Grant Build for data preparation and evaluation, or Full for training and deployment operations within your role.
Explore the featureReview spending before launch
Tools preview work that can incur provider charges. Training and deployment launches require explicit confirmation, with spending bounds for experiments.
Explore the featureKeep the work visible
Inspect the same projects, versions, and runs in the console. Tool results include relevant console links, and changes appear in the organization’s audit trail.
Explore the featureOne platform, however you work.
Use the console, CLI, and MCP against the same projects and data. MCP manages the workflow; application API keys let your software call the models you publish.
Choose your MCP client- Read
- Inspect projects, data, runs, evaluations, and models.
- Build
- Also prepare and review data, create versions, and run evaluations.
- Full
- Also manage training and deployments, with spending confirmation.
Build with Tensorant