Xmp4 — Semantic Code Knowledge For Your Stack
Connect xmp4 over MCP so your coding agent pulls SCIP-indexed source, tests, and real callers from 856 OSS libraries instead of grepping full repos.
Overview
xmp4 is an MCP server for the Build phase that serves SCIP-indexed source, tests, and caller graphs for 856 OSS libraries so coding agents stop grepping cloned repos.
What is this MCP server?
- 17 MCP tools serving 856 SCIP-indexed OSS libraries across C#, Java, TypeScript, Python, Rust, and PHP+
- Compiler-grade semantics: source, tests, call hierarchy, and who-calls-what in roughly three tool calls—no clone or grep
- Hosted Streamable HTTP at mcp.example4.ai with official MCP Registry entry ai.example4/xmp4 (live v1.2.1)
- Project benchmark docs cite about 42× fewer tokens versus GitMCP-style repo grepping for the same lookups
- Apache 2.0 OSS; connect in about 30 seconds per published connect instructions
- 856 OSS libraries indexed
- 17 MCP tools exposed
- Live server version v1.2.1 on mcp.example4.ai
What problem does it solve?
Agents waste tokens and ship wrong integrations because grepping massive library repos surfaces noise, misses real call sites, and fights stale documentation.
Who is it for?
Solo builders who lean on popular OSS packages daily and want MCP-backed, compiler-grade library context inside Claude Code or Cursor.
Skip if: Teams that only touch private monorepos with no indexed public library, or builders who refuse hosted MCP and need entirely offline grep workflows.
What do I get? / Deliverables
After you register the hosted MCP endpoint, your agent resolves APIs and callers in a few semantic tool calls with far less context than full-repo search.
- Semantic lookups across 856 indexed OSS libraries via 17 MCP tools
- Caller and hierarchy answers grounded in indexed source and tests
- Lower-context library research sessions versus full-repo grep for the same task
Recommended MCP Servers
Journey fit
Library truth lands while you implement features and wire third-party packages—the build phase is when wrong API guesses and noisy grep burn the most time and tokens. xmp4 is hosted MCP with seventeen semantic tools aimed at Claude Code, Cursor, and similar agents, which is agent-tooling rather than shipping your own backend service.
How it compares
Pre-indexed SCIP library MCP, not a clone-and-grep GitMCP workflow or a single-repo documentation skill.
Common Questions / FAQ
Who is xmp4 for?
xmp4 is for indie developers and small teams using MCP coding agents who integrate many public OSS libraries and need accurate callers and source without cloning every dependency.
When should I use xmp4?
Use xmp4 during build and integration work whenever you are implementing against unfamiliar packages, debugging how an API is really invoked, or comparing behavior to indexed tests instead of guessing from README snippets.
How do I add xmp4 to my agent?
Add the official registry server ai.example4/xmp4 or point your client at the live Streamable HTTP endpoint documented at mcp.example4.ai, following the project’s connect-instructions for your MCP host in about 30 seconds.