Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

Antigravity-mcp-semantic-search-with-TypeSafeAi

MCP server that finds relevant code with Jev Choice and Noul checks instead of grep sweeps, and audits diffs against task intent.

MCP serverprototype
View on GitHub ↗

Overview

TypeSafe Code Guard is an MCP server for AI coding assistants such as Antigravity IDE, Cursor and Claude Code. It targets the context bloat caused by keyword grep sweeps, which the README says can pull 25,000 to 40,000 irrelevant tokens into context.

typesafe_find_code is a two-stage locator: it scans the project tree and uses TypeSafe Choice and Noul evaluations to pick the most relevant file without loading contents into the LLM, then ranks line blocks in that file and returns a 15-line preview with line numbers and confidence. typesafe_audit_diff compares task intent against a git diff and flags unexpected global state changes, removed defensive checks, breaking contract changes and leftover debug statements. The README cites about $0.00004 per call. A bundled skill and rules can protect all projects in Antigravity.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich file is most relevant to this code query?—
Yes / noDoes this diff remove a defensive check / leave debug statements / break a contract?(inferred)yes · no

How it works

  1. Assistant calls typesafe_find_code with an intent
  2. Jev selects a file from the tree, then ranks line blocks
  3. Return a 15-line snippet with confidence
  4. Before commit, typesafe_audit_diff checks the diff against the task

Strengths

  • Snippets of about 150 tokens instead of 30,000+
  • Behaviour-based search rather than name guessing
  • Cheap per call, per the README

Caveats

  • Needs a TypeSafe key
  • Single-file selection may miss multi-file answers

Getting started

Python 3.10+, set TYPESAFE_API_KEY, and register typesafe-code-guard in your MCP client config.

At a glance

Type
MCP server
Status
prototype
Language
Python
Runtime
MCP server
Has code
yes

Use cases and integrations

How curated lists describe it

  • Fast semantic code search & diff sanity auditor for AI coding assistants (Antigravity, Cursor, Claude Code) powered by TypeSafe System One.

    hellogumbo/awesome-jev

Frequently asked questions

How does it find code?

Jev picks the relevant file from the tree, then ranks line blocks, without loading files into the LLM.

What does the diff audit flag?

Unexpected global mutations, removed guards, breaking contract changes and leftover debug statements.

Which assistants work?

Antigravity IDE, Cursor and Claude Code via MCP.

Related projects

  • jev-test-filter

    CLI that scores every test against a git diff with Jev and emits filter arguments for vitest, jest, node, Playwright, cargo, go and rspec.

    ProjectbetaTypeScript
  • JevTrace

    MCP server that returns only the JS/TS code a task needs by following the TypeScript compiler graph and asking Jev about relevance.

    MCP serverbetaTypeScript
  • siftr

    CLI and MCP server giving coding agents Jev-powered semantic search, focused reads and list picking in about a second for about a cent.

    BenchmarkbetaPython
  • claude-jev

    Claude Code plugin that adds Jev as an independent judge for review findings, debugging hypotheses, design options and search hits.

    Pluginbeta
  • Hunch (Kelbie)

    Semantic code search and plain-English review rules that send repository chunks to Jev and return scored source locations.

    ProjectbetaTypeScript
  • jev-kit

    Codex skill and MCP for advisory Jev decisions in Java/Maven backend work: file relevance, change risk, SQL risk and more.

    MCP serverbetaTypeScript