neuralstack
Spawn divergent-thinking whiteboards for Claude. NeuralStack lets Claude open a
thinking board on a hard, open-ended problem, hand it to you to brainstorm on,
then read your board's structure back and converge to an answer it couldn't
have produced from the bare prompt.
Install (one command)
npx neuralstack init
This will:
- Ask for your free NeuralStack API key (get one at https://thinklm.vercel.app → Settings → API key).
- Register the MCP server in your Claude config (Claude Code + Claude Desktop).
- Install the
neuralstack skill into ~/.claude/skills/.
Then restart Claude and say "use neuralstack" on any open-ended question.
Non-interactive (e.g. when an agent runs it):
npx neuralstack init --key tlm_xxx
What it registers
{
"mcpServers": {
"neuralstack": {
"command": "npx",
"args": ["-y", "neuralstack", "mcp"],
"env": { "NEURALSTACK_TOKEN": "tlm_xxx" }
}
}
}
NEURALSTACK_API_URL defaults to https://thinklm.vercel.app; set it to
http://localhost:3100 for local development.
Commands
npx neuralstack init [--key tlm_…] [--url …] | one-shot setup: MCP + skill + key |
npx neuralstack mcp | run the stdio MCP server (what the config launches) |
Tools exposed to Claude
create_board { topic } | spawn a board, returns id + preview URL |
preview_url { id } | short-lived, login-free URL for a preview pane |
wait_for_board { id } | block until the human marks the board ready |
get_board { id } | read the serialized board (tree + survivors + verdict) |
submit_answer { id, answer } | write the final answer back to the canvas |