fix(agent-runner): derive MCP allowedTools from registered mcpServers

Claude Code 2.1.116+ treats SDK `allowedTools` as a hard whitelist:
servers whose namespace isnt listed are filtered out before the agent
ever sees them, regardless of `permissionMode: bypassPermissions` or
any `permissions.allow` in settings. The static TOOL_ALLOWLIST only
contained `mcp__nanoclaw__*`, so any MCP wired via add_mcp_server (or
directly in container.json) was silently dropped.

Derive `mcp__<sanitized-name>__*` entries at the SDK call site from
the already-aggregated `this.mcpServers` map, mirroring the SDKs own
sanitization rule (chars outside [A-Za-z0-9_-] become _).

Prior diagnosis by @jsboige in #2028 (withdrawn, not upstreamed).

container/agent-runner/src/providers/claude.ts

@@ -34,7 +34,11 @@ const SDK_DISALLOWED_TOOLS = [
   'ExitWorktree',
 ];
 
-// Tool allowlist for NanoClaw agent containers
+// Tool allowlist for NanoClaw agent containers. MCP-tool entries are derived
+// at the call site from the registered `mcpServers` map so that any server
+// added via `add_mcp_server` (or wired in container.json directly) is
+// reachable to the agent — without this, the SDK's allowedTools filter
+// silently drops every MCP namespace not listed here.
 const TOOL_ALLOWLIST = [
   'Bash',
   'Read',
@@ -54,9 +58,15 @@ const TOOL_ALLOWLIST = [
   'ToolSearch',
   'Skill',
   'NotebookEdit',
-  'mcp__nanoclaw__*',
 ];
 
+// MCP server names are sanitized by the SDK when forming tool prefixes:
+// any character outside [A-Za-z0-9_-] becomes '_'. Mirror that here so our
+// allowlist patterns match what the SDK actually exposes.
+function mcpAllowPattern(serverName: string): string {
+  return `mcp__${serverName.replace(/[^a-zA-Z0-9_-]/g, '_')}__*`;
+}
+
 interface SDKUserMessage {
   type: 'user';
   message: { role: 'user'; content: string };
@@ -277,7 +287,10 @@ export class ClaudeProvider implements AgentProvider {
         resume: input.continuation,
         pathToClaudeCodeExecutable: '/pnpm/claude',
         systemPrompt: instructions ? { type: 'preset' as const, preset: 'claude_code' as const, append: instructions } : undefined,
-        allowedTools: TOOL_ALLOWLIST,
+        allowedTools: [
+          ...TOOL_ALLOWLIST,
+          ...Object.keys(this.mcpServers).map(mcpAllowPattern),
+        ],
         disallowedTools: SDK_DISALLOWED_TOOLS,
         env: this.env,
         permissionMode: 'bypassPermissions',