Live demo · Python in your browser · simulated data

ephys-mcp tool console

Call the server's analysis tools the way an LLM would, on a simulated motor cortex

The page loads Python with NumPy and SciPy, unpacks the repo's v0.5.1 package and calls its tool functions directly. 20 of its 24 MCP tools run here; spike sorting, DANDI file listing and the two live-stream tools are listed but switched off, and search_datasets only returns its built-in offline list. The data comes from the server's synthetic source: cosine-tuned Poisson units with known spikes and cursor velocity. No language model is involved. You press the buttons, and every number is computed here.

  1. list_sources
  2. open_session
  3. detect_spikes
  4. fit_decoder
  5. get_psth
  6. fit_latent_factors

About 25 MB from cdn.jsdelivr.net (Pyodide 0.29.5, NumPy, SciPy), cached after the first visit. Nothing loads until you press Start.