# Ragsharp MCP Code Graph

> Run the ragsharp MCP JSON-RPC server and issue graph commands (doctor/index/update/query/export) for Codex/agent integrations. Triggers: mcp, json-rpc, code graph, references, index, query, export.

- Skill: `managedcode/ragsharp-mcp-code-graph` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add managedcode/ragsharp-mcp-code-graph`
- Raw SKILL.md: https://api.skillmd.com/api/skills/managedcode/ragsharp-mcp-code-graph/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: managedcode (https://skillmd.com/u/managedcode)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/managedcode/ragsharp-mcp-code-graph

---

## Steps
1. Start the MCP server: `ragsharp-mcp` (reads stdin, writes stdout).
2. Send JSON-RPC 2.0 requests, e.g. `{"jsonrpc":"2.0","id":1,"method":"doctor","params":{"root":"."}}`.
3. For indexing: `{"jsonrpc":"2.0","id":2,"method":"index","params":{"root":".","db": ".ragsharp/graph/index.db","state": ".ragsharp/graph/state.json"}}`.
4. For querying: `{"jsonrpc":"2.0","id":3,"method":"query","params":{"db": ".ragsharp/graph/index.db","type":"symbols","symbol":"Greeter","edgeKind":null,"limit":50}}`.
5. For export: `{"jsonrpc":"2.0","id":4,"method":"export","params":{"db": ".ragsharp/graph/index.db","format":"dot","output": ".ragsharp/graph/graph.dot"}}`.

## Expected Results
- Server responds with JSON-RPC 2.0 responses containing `result` or `error`.
- Indexing returns counts of nodes/edges and writes `.ragsharp/graph/index.db` and `state.json`.
- Query returns `nodes` and `edges` matching filters (`type`, `symbol`, `kind`, `document`, `edgeKind`, `limit`, `contextLines`).
- Export writes DOT or GEXF to the requested path.