diff --git a/src/mcp/index.ts b/src/mcp/index.ts
index bc3552ae..e516631a 100644
--- a/src/mcp/index.ts
+++ b/src/mcp/index.ts
@@ -19,6 +19,7 @@ import * as path from 'path';
 import CodeGraph, { findNearestCodeGraphRoot } from '../index';
 import { StdioTransport, JsonRpcRequest, JsonRpcNotification, ErrorCodes } from './transport';
 import { tools, ToolHandler } from './tools';
+import { SERVER_INSTRUCTIONS } from './server-instructions';
 
 /**
  * Convert a file:// URI to a filesystem path.
@@ -268,13 +269,17 @@ export class MCPServer {
     // Try to initialize the default project (non-fatal if it fails)
     await this.tryInitializeDefault(projectPath);
 
-    // We accept the client's protocol version but respond with our supported version
+    // We accept the client's protocol version but respond with our supported version.
+    // The `instructions` field is surfaced by MCP clients in the agent's system
+    // prompt automatically — it's the right place for the universal tool-selection
+    // playbook, ahead of individual tool descriptions.
     this.transport.sendResult(request.id, {
       protocolVersion: PROTOCOL_VERSION,
       capabilities: {
         tools: {},
       },
       serverInfo: SERVER_INFO,
+      instructions: SERVER_INSTRUCTIONS,
     });
   }
 
diff --git a/src/mcp/server-instructions.ts b/src/mcp/server-instructions.ts
new file mode 100644
index 00000000..0c715ea8
--- /dev/null
+++ b/src/mcp/server-instructions.ts
@@ -0,0 +1,55 @@
+/**
+ * Server-level instructions emitted in the MCP `initialize` response.
+ *
+ * MCP clients (Claude Code, Cursor, opencode, LangChain, OpenAI Agent
+ * SDK, …) surface this text in the agent's system prompt automatically,
+ * giving the agent a high-level playbook for the codegraph toolset
+ * before it sees individual tool descriptions.
+ *
+ * Goals when editing this:
+ *   - Tool selection by intent (which tool for which question)
+ *   - Common chains (refactor planning = X then Y)
+ *   - Anti-patterns (don't grep when codegraph_search is faster)
+ *
+ * Keep it tight. The agent reads this every session — long instructions
+ * burn tokens. Reference only tools that exist on `main`; gate any
+ * conditional tools behind feature checks if/when they ship.
+ */
+export const SERVER_INSTRUCTIONS = `# Codegraph — code intelligence over an indexed knowledge graph
+
+Codegraph is a SQLite knowledge graph of every symbol, edge, and file
+in the workspace. Reads are sub-millisecond; the index lags writes by
+about a second through the file watcher. Consult it BEFORE writing or
+editing code, not during.
+
+## Tool selection by intent
+
+- **"What is the symbol named X?"** → \`codegraph_search\`
+- **"What's the deal with this task / feature / area?"** → \`codegraph_context\` (PRIMARY — composes search + node + callers + callees in one call)
+- **"What calls this?"** → \`codegraph_callers\`
+- **"What does this call?"** → \`codegraph_callees\`
+- **"What would changing this break?"** → \`codegraph_impact\`
+- **"Show me this symbol's source / signature / docstring."** → \`codegraph_node\`
+- **"Survey an unfamiliar topic / pattern / module."** → \`codegraph_explore\` (heavier; deep dive)
+- **"What's in directory X?"** → \`codegraph_files\`
+- **"Is the index ready / what's its size?"** → \`codegraph_status\`
+
+## Common chains
+
+- **Onboarding**: \`codegraph_context\` first. If still unclear, \`codegraph_explore\` for breadth, then \`codegraph_node\` on specific symbols.
+- **Refactor planning**: \`codegraph_search\` → \`codegraph_callers\` → \`codegraph_impact\`. The blast-radius answer comes from impact, not from walking callers manually.
+- **Debugging a regression**: \`codegraph_callers\` of the suspected symbol; widen with \`codegraph_impact\` if an unexpected call appears.
+
+## Anti-patterns
+
+- **Don't grep first** when looking up a symbol by name — \`codegraph_search\` is faster and returns kind + location + signature.
+- **Don't chain \`codegraph_search\` + \`codegraph_node\`** when you just want context — \`codegraph_context\` is one round-trip.
+- **Don't use \`codegraph_explore\` for narrow questions** — it's a multi-call deep dive, expensive in tokens. Save it for genuine "I'm new here" surveys.
+- **Don't query the index immediately after editing a file** — the watcher needs ~500ms to debounce + sync. Wait for the next turn.
+
+## Limitations
+
+- Index lags file writes by ~1 second.
+- Cross-file resolution is best-effort name matching; ambiguous calls may return multiple candidates.
+- No live correctness validation — that's still the TypeScript compiler / test suite / linter's job. Codegraph supplements those with structural context they don't have.
+`;