TypeSafe — Claude Code MCP server for data judgments
TypeSafe is an MCP server for Claude Code that scores claims, choices, and ordered scales. It can handle one item or batches of up to 1000 rows.
MIT ★ 5 updated 2026-10-29
claude mcp add typesafe -e TYPESAFE_API_KEY_FILE=<TYPESAFE_API_KEY_FILE> -- npx -y @foreverworks/mcp-typesafeWhat TypeSafe does
TypeSafe sends questions to Jev and returns probabilities, a selected option, or a score on an ordered scale. The ask tool handles a single state; ask_batch processes a set and can apply thresholds or filters such as where, top, and counts.
The server needs a TypeSafe account and an API key stored in a file named through TYPESAFE_API_KEY_FILE. Calls are billed by input tokens, and the published package requires Node 22.19 or newer. Batch input files are read from MCP_FILE_ROOTS; full results can be saved as JSONL.
What it helps with
- estimate whether a claim holds for an item
- choose a category from a set of options
- score up to 1000 rows on an ordered scale
- filter batch results and count categories
Example prompts for the agent
- Estimate the chance that this record meets the criteria and include Jev’s confidence.
- Classify the entries in this file and save the full results as JSONL.
- Score each row on this scale and show the ones above the threshold.
How to install TypeSafe
Platform CLI
claude mcp add typesafe -e TYPESAFE_API_KEY_FILE=<TYPESAFE_API_KEY_FILE> -- npx -y @foreverworks/mcp-typesafeReplace <…> with your own values
Platform CLI
codex mcp add typesafe --env TYPESAFE_API_KEY_FILE=<TYPESAFE_API_KEY_FILE> -- npx -y @foreverworks/mcp-typesafeReplace <…> with your own values
Config file
{
"mcpServers": {
"typesafe": {
"command": "npx",
"args": [
"-y",
"@foreverworks/mcp-typesafe"
],
"env": {
"TYPESAFE_API_KEY_FILE": "<TYPESAFE_API_KEY_FILE>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"typesafe": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@foreverworks/mcp-typesafe"
],
"env": {
"TYPESAFE_API_KEY_FILE": "<TYPESAFE_API_KEY_FILE>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e TYPESAFE_API_KEY_FILE=<TYPESAFE_API_KEY_FILE> typesafe npx -y @foreverworks/mcp-typesafeReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about TypeSafe
How do I install TypeSafe in Claude Code?
Run claude mcp add typesafe -e TYPESAFE_API_KEY_FILE=<TYPESAFE_API_KEY_FILE> -- npx -y @foreverworks/mcp-typesafe. Set the placeholder to the path of your TypeSafe API key file.
What does TypeSafe do?
It sends questions about a single item or a batch of up to 1000 rows to Jev. Responses include a probability, a choice among options, or a score on an ordered scale.
What does TypeSafe need to run?
You need a TypeSafe account, an API key stored in a file referenced by TYPESAFE_API_KEY_FILE, and Node 22.19 or newer for the published package. Calls are billed by input tokens.