Add vexp (local-first context engine) and vexp-swe-bench (open benchmark harness)

[nicolalessi]

What

Adding two related resources:

1. vexp - A local-first context engine for Claude Code (and 11 other agents)
   that serves ranked context via MCP instead of letting the agent read entire files.
   Runs 100% locally using tree-sitter + SQLite. Auto-configures CLAUDE.md and hooks.

2. vexp-swe-bench - An open-source benchmark harness for evaluating coding agents
   on SWE-bench Verified. Agent-agnostic (adapter interface for adding new agents).
   Includes full evaluation logs, result JSONLs, and comparison charts.

Links

• VS Code extension: https://marketplace.visualstudio.com/items?itemName=Vexp.vexp-vscode
• CLI: https://www.npmjs.com/package/vexp-cli
• Benchmark repo: https://github.com/Vexp-ai/vexp-swe-bench
• Benchmark results: https://vexp.dev/benchmark
• Docs: https://vexp.dev/docs

Why it fits this list

• Works specifically with Claude Code (auto-generates CLAUDE.md, installs PreToolUse hooks,
  configures MCP)
• The benchmark harness is open source (MIT) and designed for the community to add
  new agent adapters
• Zero network calls, all local processing