generated: '2026-07-20'
method: searched
status: published
source: https://docs.priorlabs.ai/agentic/tool-use.md
server:
  name: tabpfn
  transport: http
  url: https://api.priorlabs.ai/mcp/server
  auth:
    methods:
    - oauth
    - bearer
    note: Claude.ai / Claude Desktop / ChatGPT / Codex CLI / Cursor use OAuth; manual config accepts an
      Authorization Bearer API key from ux.priorlabs.ai.
  install:
    claude_code: claude mcp add --transport http tabpfn https://api.priorlabs.ai/mcp/server
docs:
  setup: https://docs.priorlabs.ai/agentic/setup-guide.md
  tools: https://docs.priorlabs.ai/agentic/tool-use.md
  overview: https://docs.priorlabs.ai/agentic/mcp.md
tools:
- name: upload_dataset
  description: Obtain a secure upload URL for training or test data; returns a dataset_id valid for 60
    minutes.
- name: fit_and_predict_from_dataset
  description: Fit the TabPFN model on pre-uploaded data and generate predictions.
- name: predict_from_dataset
  description: Run predictions with an existing fitted model against a new uploaded test dataset.
- name: fit_and_predict_inline
  description: Train a model directly from inline data arrays and immediately return predictions plus
    a model ID.
- name: predict_inline
  description: Generate predictions from a previously fitted model given a model ID and inline test features.
deployment:
  mode: remote
  endpoint: https://api.priorlabs.ai/mcp/server
  verified: probed
  probe: gated
  checked: '2026-08-12'
  source: catalog MCP census