Fix cron reminder phrasing and improve MCP tool visibility/status on long-running instances

docs/cron.md

@@ -36,6 +36,8 @@ This is the default for the cron tool.
 
 When the job fires, PicoClaw sends the saved message back through the agent loop as a new agent turn. Use this for scheduled work that may need reasoning, tools, or a generated reply.
 
+Because the saved `message` is replayed as a new user-style input, write it from the user's perspective or as direct instructions to the agent. Prefer wording such as `check the repo every hour and tell me if there is a new release` over third-person wording such as `check the repo and notify the user`.
+
 ### `deliver: true`
 
 When the job fires, PicoClaw publishes the saved message directly to the target channel and recipient without agent processing.
@@ -50,6 +52,23 @@ For command jobs, `deliver` is forced to `false` when the job is created. The sa
 
 The current CLI `picoclaw cron add` command does not expose a `command` flag.
 
+## Writing Job Messages
+
+For normal cron jobs without `command`, the saved `payload.message` becomes the next input sent to the agent when the job fires. In practice, that means the job message should read like something the user would say to the agent.
+
+Recommended style:
+
+- Use first-person or direct-address wording such as `tell me`, `remind me`, `reply in Chinese`, `do not reply if nothing changed`
+- Be explicit about the quiet case if needed, for example `If there is no update, do not reply`
+- Avoid third-person wording such as `notify the user`, because the model may continue replying in third person
+
+Examples:
+
+```text
+Good: Check gdsfactory/gdsfactory every hour. If there is a new release, tell me in Chinese and summarize the changes. If nothing changed, do not reply.
+Bad: Check gdsfactory/gdsfactory every hour. If there is a new release, notify the user and summarize the changes.
+```
+
 ## Config and Security Gates
 
 ### `tools.cron`

docs/tools_configuration.md

@@ -252,6 +252,8 @@ The cron tool is used for scheduling periodic tasks.
 | `allow_command`        | bool | true    | Allow command jobs without extra confirmation  |
 | `exec_timeout_minutes` | int  | 5       | Execution timeout in minutes, 0 means no limit |
 
+For normal cron jobs without `command`, the saved `message` is later replayed into the agent loop as a new user-style message. Write it from the user's perspective, for example `check the repo every hour and tell me if there is a new release`, not `notify the user`.
+
 For schedule types, execution modes (`deliver`, agent turn, and command jobs), persistence, and the current command-security gates, see [Scheduled Tasks and Cron Jobs](cron.md).
 
 ## MCP Tool

docs/zh/tools_configuration.md

@@ -234,6 +234,8 @@ Cron 工具用于调度周期性任务。
 | `exec_timeout_minutes` | int  | 5      | 执行超时时间（分钟），0 表示无限制  |
 | `allow_command`        | bool | false  | 允许 cron 任务执行 shell 命令       |
 
+对于不带 `command` 的普通 cron 任务，保存下来的 `message` 会在任务触发时重新作为一条新的“用户消息”送回 agent。编写时应使用用户视角/直接对 agent 说话的口吻，例如 `每小时检查仓库更新，如果有新版本告诉我`，而不是 `如果有更新就通知用户`。
+
 ## MCP 工具
 
 MCP 工具支持与外部 Model Context Protocol 服务器集成。

pkg/agent/loop.go

@@ -1033,6 +1033,47 @@ func (al *AgentLoop) ReloadProviderAndConfig(
 	// Ensure shared tools are re-registered on the new registry
 	registerSharedTools(al, cfg, al.bus, registry, provider)
 
+	var (
+		newMCPManager mcpController
+		mcpSummary    mcpRegistrationSummary
+	)
+	if cfg.Tools.IsToolEnabled("mcp") && countEnabledMCPServers(cfg.Tools.MCP.Servers) > 0 {
+		newMCPManager = newMCPController()
+
+		workspacePath := cfg.WorkspacePath()
+		if defaultAgent := registry.GetDefaultAgent(); defaultAgent != nil && defaultAgent.Workspace != "" {
+			workspacePath = defaultAgent.Workspace
+		}
+
+		if err := newMCPManager.LoadFromMCPConfig(ctx, cfg.Tools.MCP, workspacePath); err != nil {
+			if closeErr := newMCPManager.Close(); closeErr != nil {
+				logger.ErrorCF("agent", "Failed to close MCP manager",
+					map[string]any{
+						"error": closeErr.Error(),
+					})
+			}
+			return fmt.Errorf("failed to initialize MCP during reload: %w", err)
+		}
+
+		mcpSummary = registerMCPToolsOnRegistry(registry, cfg, newMCPManager, newMCPManager.GetServers())
+		if err := registerMCPDiscoveryToolsOnRegistry(registry, cfg); err != nil {
+			if closeErr := newMCPManager.Close(); closeErr != nil {
+				logger.ErrorCF("agent", "Failed to close MCP manager",
+					map[string]any{
+						"error": closeErr.Error(),
+					})
+			}
+			return fmt.Errorf("failed to restore MCP discovery tools during reload: %w", err)
+		}
+	}
+	if al.mediaStore != nil {
+		for _, agentID := range registry.ListAgentIDs() {
+			if agent, ok := registry.GetAgent(agentID); ok {
+				agent.Tools.SetMediaStore(al.mediaStore)
+			}
+		}
+	}
+
 	// Atomically swap the config and registry under write lock
 	// This ensures readers see a consistent pair
 	al.mu.Lock()
@@ -1054,6 +1095,7 @@ func (al *AgentLoop) ReloadProviderAndConfig(
 
 	al.mu.Unlock()
 
+	oldMCPManager := al.mcp.replaceForReload(newMCPManager, newMCPManager != nil)
 	al.hookRuntime.reset(al)
 	configureHookManagerFromConfig(al.hooks, cfg)
 
@@ -1074,10 +1116,21 @@ func (al *AgentLoop) ReloadProviderAndConfig(
 			}
 		}
 	}
+	if oldMCPManager != nil {
+		if err := oldMCPManager.Close(); err != nil {
+			logger.ErrorCF("agent", "Failed to close previous MCP manager",
+				map[string]any{
+					"error": err.Error(),
+				})
+		}
+	}
 
 	logger.InfoCF("agent", "Provider and config reloaded successfully",
 		map[string]any{
-			"model": cfg.Agents.Defaults.GetModelName(),
+			"model":                   cfg.Agents.Defaults.GetModelName(),
+			"mcp_server_count":        mcpSummary.serverCount,
+			"mcp_unique_tools":        mcpSummary.uniqueTools,
+			"mcp_total_registrations": mcpSummary.totalRegistrations,
 		})
 
 	return nil
@@ -3454,6 +3507,9 @@ func (al *AgentLoop) buildCommandsRuntime(agent *AgentInstance, opts *processOpt
 		Config:          cfg,
 		ListAgentIDs:    registry.ListAgentIDs,
 		ListDefinitions: al.cmdRegistry.Definitions,
+		GetMCPStatus: func() string {
+			return formatMCPStatus(cfg, al.mcp.statusSnapshot())
+		},
 		GetEnabledChannels: func() []string {
 			if al.channelManager == nil {
 				return nil