Skip to content

v2.2.0: evaluate on your own code

Choose a tag to compare

View all tags
@jaggernaut007 jaggernaut007 released this 03 Nov 14:19
· 23 commits to main since this release
v2.2.0
7f32aca
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Measure Nexus-MCP on your own code before you adopt it.

uvx nexus-mcp-ci eval            # the current folder: free, offline, no LLM
uvx nexus-mcp-ci eval --example  # the Nexus-MCP source

New

  • nexus-mcp-ci eval: questions made from your own code, Grep and Read against one Nexus-MCP search. It prints a verdict ("Strong fit", "Likely fit" or "Weak fit"). It calls no LLM and sends nothing. See docs/EVALUATE.md.
  • eval --agent claude|codex (paid, opt-in): real agent runs with your own CLI login, a cost estimate first, and a --max-usd stop. Codex support is experimental.
  • eval doctor, eval list|show, eval share: share is optional, shows the exact payload, and posts nothing without a yes. There is no telemetry.
  • nexus-mcp-ci data where|clear|export-memories: see and remove every stored file. See docs/PRIVACY.md.
  • NEXUS_OFFLINE=1: no network after the first model download.
  • Published benchmarks come from one file, with a benchmarks page (HTML, JSON, CSV).
  • Listed in the official MCP Registry as io.github.jaggernaut007/nexus-mcp.

Quick-tier results (2026-11-03)

Project Files Verdict Median tokens: Grep and Read → Nexus-MCP Right file first
Chromium base/ 3,638 Strong fit 17,998 → 950 40% → 70%
pallets/click 91 Strong fit 24,774 → 1,301 27% → 97%
Nexus-MCP source 59 Likely fit 5,385 → 1,234 70% → 93%

The token numbers are a proxy (characters divided by 4), not a bill.

Compatibility

No breaking change. The 8 MCP tools and their schemas do not change. Full list: CHANGELOG.

Assets 2
Loading