# xmp4 โ Semantic Code Intelligence [Health: Active]
**Category:** ๐ป Developer Tools
**Repository:** https://github.com/0ics-srls/lsai-xmp4.public
**GitHub Stars:** 0
**Views:** 0
**Installs:** 0
**Upvotes:** 0
**Directory Page:** https://allmcps.com/mcp/xmp4-semantic-code-intelligence
## Description
SCIP-backed semantic code intelligence over 800+ pre-indexed OSS libraries. Hosted, free.
## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: high). Install path detected from listing signals. Add as a URL/SSE server in your client:
```json
"mcpServers": {
"xmp4-semantic-code-intelligence": {
"url": "https://img.shields.io/badge/mcp.example4.ai-live%20v1.2.1-22c55e?style=flat-square)](https://mcp.example4.ai/mcp"
}
}
```
## Documentation & README
# xmp4
### Stop grepping library source.
### Your AI gets the compiler's view of **856 OSS libraries** via MCP.
[](https://mcp.example4.ai/mcp)
[](https://registry.modelcontextprotocol.io/v0/servers?search=xmp4)
[](LICENSE)
[](https://modelcontextprotocol.io)
[](#language-coverage)
[](docs/benchmarks/WHITEPAPER.md)
**Real callers. Real source. Real hierarchy. In 3 tool calls.**
[**โ Landing page**](https://example4.ai) ยท [**โ Benchmark whitepaper**](docs/benchmarks/WHITEPAPER.md) ยท [**โ Connect in 30 seconds**](docs/connect-instructions.md)
---
## The 30-second pitch
Your AI coding agent is burning tokens grepping OSS libraries it will barely use. **xmp4** is a hosted MCP server that pre-indexes 856 popular open-source libraries with **SCIP** โ the semantic code format Sourcegraph uses โ and serves them through 17 tools. No clone. No grep. No false positives.
```text
ASK: "Who calls Flask.wsgi_app in the flask repo and what does it do?"
with grep + local clone:
git clone flask/flask ~40 MB, ~2 min
grep -rn "wsgi_app" . 200+ matches, mostly noise
cat src/flask/app.py | sed ... read 1000+ lines to find the body
filter false positives model spends tokens deciding what's real
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
total: ~15,000 tokens + disk + wall time
with xmp4:
xmp4_info(symbol_name="Flask", file_path="src/flask/app.py") โ signature, 20 tok
xmp4_source(symbol_name="wsgi_app", file_path="src/flask/app.py") โ body, 180 tok
xmp4_callers(symbol_name="wsgi_app", file_path="src/flask/app.py") โ 1 caller, 50 tok
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
total: ~250 tokens
xmp4 is 60ร cheaper here โ and every result is SCIP-resolved, not text-matched.
```
## The measured numbers (4 big OSS libs ยท [reproducible](docs/benchmarks/WHITEPAPER.md))
Same realistic question on spring-boot ยท tokio ยท django ยท efcore: *"give me the signature, body, and real callers of X."*
| | **xmp4** | grep + clone | GitMCP | Context7 |
|---|---:|---:|---:|---:|
| Total tokens (same question) | **1 558** | 2 978 | **65 629** | โ |
| vs xmp4 | 1ร | 1.9ร more | **42ร more** | can't answer |
| Returns real source body? | โ
| โ
noisy | โ file paths only | โ curated docs only |
| Semantic callers? | โ
| โ | โ | โ |
| Type hierarchy? | โ
| โ | โ | โ |
| Setup cost | 0 | GBs of clone | 0 | 0 |
> GitMCP and Context7 look cheaper **per call** because they return less. To reach the same answer, GitMCP balloons to 42ร more tokens โ and still can't produce the semantic caller list. Context7 can't at any cost. [**Full whitepaper with Python harness โ**](docs/benchmarks/WHITEPAPER.md)
## Connect in 30 seconds
```jsonc
// Claude Code / Cursor / Claude Desktop โ project `.mcp.json` or client config
{
"mcpServers": {
"xmp4": {
"type": "http",
"url": "https://mcp.example4.ai/mcp"
}
}
}
```
(The ready-to-paste config also lives at [`.mcp.json`](.mcp.json) in this repo.)
### Teach Claude how to use xmp4 (optional but recommended)
Install the xmp4 skill once per version โ Claude will pick the cheapest tool path automatically (tests_for + view over grep):
```bash
# Claude Code
mkdir -p ~/.claude/skills/xmp4 && \
curl -sfL https://example4.ai/xmp4-skill.md -o ~/.claude/skills/xmp4/SKILL.md
# Other clients: just tell Claude to read the URL when using xmp4 tools
# https://example4.ai/xmp4-skill.md
```
Restart your client. Then try (every step verified live 2026-04-24):
> *"Using xmp4, find the Flask class in `flask/Flask` and list its usages."*
You should see `Type Flask src/flask/app.py:81` and 165 usages across 33 result pages. One semantic call per question. Zero grep loops.
**Setup for Cursor ยท Claude Desktop ยท Continue ยท Windsurf** โ [`docs/connect-instructions.md`](docs/connect-instructions.md)
## The 17 tools
[**Full reference with live examples**](docs/tool-reference.md)
**Semantic core** (where the value lives)
`xmp4_projects` ยท `xmp4_search` ยท `xmp4_info` ยท `xmp4_usages` ยท `xmp4_callers` ยท `xmp4_callees` ยท `xmp4_hierarchy` ยท `xmp4_outline` ยท `xmp4_source` ยท `xmp4_tests_for` ยท `xmp4_deps` ยท `xmp4_symbol_at`
**Convenience**
`xmp4_view` (raw file excerpt by line range) ยท `xmp4_grep` (server-side regex when semantics isn't enough)
**Meta**
`xmp4_guide` (returns a **versioned skill pointer** to `https://example4.ai/xmp4-skill.md` โ fetch once per version and save as a local Claude Code skill; embeds a minimal cheatsheet as offline fallback) ยท `xmp4_server` (version + stats)
## Language coverage
**Tier 1** โ full coverage contract
`C#` ยท `TypeScript` ยท `Python` ยท `Java` ยท `Rust` ยท `PHP`
**Tier 2** โ best-effort, documented quirks
`Go` ยท `JavaScript` ยท `Dart` ยท `Ruby` ยท `C++`
Every known limitation โ empty `hierarchy.base` on TS/Rust/Java/PHP, Python cross-module `usages` under-count, C# explicit-interface-impl dotted-name behaviour โ is listed verbatim in [`docs/tiers-and-quirks.md`](docs/tiers-and-quirks.md). We'd rather set expectations correctly than have a reviewer find a gap and assume the whole thing is inflated.
## Coverage grows by demand, not by guesswork
The index currently holds 856 repositories / 15 921 SCIP-indexed projects. We add new libraries based on two signals, combined:
1. **Aggregate query logs** โ symbol names and project filters, no PII, no user code. If many AI agents search for a library we don't have, we see it.
2. **Your request** โ [file a repo-request issue](../../issues/new?template=request-repo.yml) with the GitHub URL, the language, and one concrete query you want to run. A single user request + downstream query demand almost always means **indexed within days**.
A public `/stats/top-missing` endpoint is planned โ full transparency on what drives the growth loop.
## Privacy โ short version
- โ **We log**: aggregate query counts (symbol/project/tool names, coarse timestamps) to grow the index by demand.
- โ **We don't log**: the contents of your codebase ยท personal identifiers ยท request bodies beyond declared tool parameters.
- Standard nginx access logs kept 7 days for abuse prevention, then purged. Not joined with query tallies.
Full detail โ [`docs/privacy.md`](docs/privacy.md).
## Status
- ๐ข **Live** โ `mcp.example4.ai` v1.2.1 ยท 856 repos ยท 15 921 projects ยท 17 tools ยท 11 languages
- ๐ข **Benchmark published** โ [reproducible whitepaper with Python harness](docs/benchmarks/WHITEPAPER.md)
- ๐ข **Listed on the Official MCP Registry** as [`ai.example4/xmp4`](https://registry.modelcontextprotocol.io/v0/servers?search=xmp4) (DNS-authed on `example4.ai`)
- ๐ข **Smithery** โ [smithery.ai/servers/0ics-srl/xmp4](https://smithery.ai/servers/0ics-srl/xmp4)
- ๐ข **Cursor Directory** โ [cursor.directory/plugins/lsai-xmp4public](https://cursor.directory/plugins/lsai-xmp4public)
- ๐ง **More registry submissions in flight** โ MCP.so, mcpservers.org, partnerships@github.com, awesome-lists (PRs open on punkpeye/jaw9c/appcypher)
- ๐ง **Demand-driven growth loop** โ in progress
## What's in this repository
| Path | Purpose |
|---|---|
| [`docs/connect-instructions.md`](docs/connect-instructions.md) | 5 MCP clients, proof-of-life sequence, troubleshooting |
| [`docs/tool-reference.md`](docs/tool-reference.md) | 17 tools with live-verified examples and workflow rules |
| [`docs/tiers-and-quirks.md`](docs/tiers-and-quirks.md) | Language tier matrix + every known limitation, verbatim |
| [`docs/privacy.md`](docs/privacy.md) | What we log, what we don't, GDPR contact |
| [`docs/request-repo.md`](docs/request-repo.md) | How the demand-driven queue actually works |
| [`.mcp.json`](.mcp.json) | Ready-to-paste MCP client config (`type: http`, URL) |
| [`skills/xmp4/SKILL.md`](skills/xmp4/SKILL.md) | Claude Code skill โ workflow, cost budget, grep policy, common mistakes |
| [`html/xmp4-skill.md`](html/xmp4-skill.md) | Same skill, served publicly at `https://example4.ai/xmp4-skill.md` |
| [`server.json`](server.json) | Official MCP Registry manifest (DNS-authed `ai.example4/xmp4`) |
| [`glama.json`](glama.json) | Glama catalog auto-index hook |
| [`.github/ISSUE_TEMPLATE/`](.github/ISSUE_TEMPLATE/) | Bug ยท feature-request ยท **request-repo** templates |
## Related
- [**LSAI protocol**](https://github.com/LadislavSopko/lsai-protocol) โ open spec for semantic code intelligence in AI agents.
- [**SCIP**](https://github.com/sourcegraph/scip) โ the semantic code format xmp4 is built on (Sourcegraph-developed, BSD-3).
- [**Model Context Protocol**](https://modelcontextprotocol.io) โ the open transport spec xmp4 speaks.
## License
Apache 2.0 for this documentation repository โ see [LICENSE](LICENSE).
The hosted `mcp.example4.ai` endpoint is free for personal and commercial use (TOS link pending).
Commercial licensing or self-hosted deployment enquiries โ open a GitHub issue labelled `commercial` on this repo.
---
Made with semantic intelligence instead of grep.
[SCIP](https://github.com/sourcegraph/scip) ยท [MCP](https://modelcontextprotocol.io) ยท [LSAI](https://github.com/LadislavSopko/lsai-protocol)