xmp4

Install once. Your AI gets compiler-grade intelligence on the codebase you're actively developing — callers, impact, hierarchy, rename.

- "Who calls myProcessOrdermethod?" → 15 callers, exact lines
- "What breaks if I changePublishAsync?" → 27 tests, risk: MEDIUM
- "Show me the inheritance chain ofBaseService" → full hierarchy
- "RenameuserIdsafely across the workspace" → done, all references updated

14 tools· 9 languages · runs on your machine · zero cloud

One line in your MCP config. Your AI navigates 900+ OSS libraries with type-resolved precision — callers, source, hierarchy, tests.

- "How doesMongoClient.connect()work?" → full source + 26 callers
- "Show me who callsapp.routein Flask" → typed call sites
- "What's the hierarchy ofMatrixin Eigen?" → compiler-resolved
- "Find tests forS3Client.PutObject" → real test files

17 tools· 12 languages · 900+ libraries ·15,852 projects· free

LSAI tells you"your PublishAsync calls GetRequiredService".
xmp4 shows youhow GetRequiredService is implemented in dotnet/runtime.
Your AI never guesses. Itknows.

Developer asks: "What does my PublishAsync call, and how does that dependency work?"

lsai_callees("PublishAsync") → CreateScope() — PublishService.cs:22 → GetRequiredService<LsaiDbContext> — PublishService.cs:23 → Where<PublishedProjectEntity> — PublishService.cs:74

↓ the AI now knows which library API to investigate

xmp4 Step 2 — How does that LIBRARY API work?

xmp4_source("GetRequiredService", project="dotnet/runtime") → public interface ISupportRequiredService { object GetRequiredService(Type serviceType); } // Implementation in ServiceProviderServiceExtensions.cs: // Throws InvalidOperationException if service not registered

LSAI Step 3 — What's the risk of changing it?

lsai_impact("PublishAsync") → MEDIUM risk — 15 references, 15 callers, 27 tests affected Tests: PublishServiceTests, V32SpecComplianceTests, PublishSearchVectorTests... All tests named, all lines numbered.

xmp4 vs alternatives — Task: "Give me the signature, body, and real callers of X"

Measured on 4 tier-1 libraries: spring-boot, tokio, django, efcore

● both LSAI + xmp4 · ● LSAI only · ● xmp4 only · ◐ best-effort · — not yet

What's the difference between LSAI and xmp4?

LSAI runs on your machine and analyzes your code in real time via LSP servers (Roslyn, jdtls, rust-analyzer, etc.). xmp4 is a cloud service with 900+ pre-indexed OSS libraries via SCIP. Use LSAI for your codebase, xmp4 for third-party libraries. They share the same MCP protocol — your AI uses both seamlessly.

Each works standalone. xmp4 alone gives your AI deep knowledge of OSS libraries. LSAI alone gives it deep knowledge of your code. Together, the AI has zero blind spots — it can trace a call from your code into a library's internals and back, all with compiler-grade precision.

No. xmp4 consumes SCIP (an open Apache 2.0 indexing format from Sourcegraph) but is an independent product by 0ics srl. LSAI uses standard LSP servers. Neither product is affiliated with Sourcegraph.

Never. LSAI runs 100% locally — your code stays on your machine. xmp4 is read-only on public OSS libraries that we indexed. The two communicate with your AI agent via MCP, not with each other.

How is xmp4 different from Context7 or GitMCP?

Context7 serves documentation snippets — no source code, no call graph. GitMCP does text search on GitHub — no type info. xmp4 has the compiler's view: typed callers, real source, class hierarchies. Same answer, 42× fewer tokens (measured).

Any MCP-capable client: Claude Code, Claude Desktop, Cursor, VS Code (with Copilot MCP), Windsurf, Zed, and more. Both LSAI and xmp4 speak standard MCP.

Open a request on GitHub. Popular OSS libraries are added within a day.

# Install (Linux / macOS / WSL) curl -fsSL https://github.com/0ics-srls/Zerox.Lsai.Public/releases/latest/download/install.sh | bash # Add to .mcp.json (once) { "mcpServers": { "lsai": { "command": "~/.lsai/run", "args": ["--stdio"] } } } # Done. cd into your project, launch claude.

One line in your MCP config. No install. No API key.

# Add to .mcp.json (Claude Desktop, Cursor, VS Code) { "mcpServers": { "xmp4": { "type": "http", "url": "https://mcp.example4.ai/mcp" } } } # Optional: install the xmp4 skill for Claude Code mkdir -p ~/.claude/skills/xmp4 && curl -sfL \ https://example4.ai/xmp4-skill.md \ -o ~/.claude/skills/xmp4/SKILL.md

LSAI docs →·xmp4 docs →·xmp4 skill →

Integrates with the Gerrit code review system to review code changes and details.