Update documentation

Author: Your Name

cecli/website/docs/config/agent-mode.md

@@ -52,6 +52,7 @@ Agent Mode uses a centralized local tool registry that manages all available too
 - **Git Tools**: `GitDiff`, `GitLog`, `GitShow`, `GitStatus`
 - **Utility Tools**: `UpdateTodoList`, `UndoChange`, `Finished`
 - **Skill Management**: `LoadSkill`, `RemoveSkill`
+- **Sub-Agent Tools**: `Delegate` - Delegate sub-tasks to specialized sub-agents
 
 #### Enhanced Context Management
 
@@ -154,6 +155,14 @@ agent-config:
   tools_excludelist: ["command", "commandinteractive"]  # Optional: Blacklist of tools
   tools_paths: ["./custom-tools", "~/my-tools"]  # Optional: Directories or files containing custom tools
 
+  # Server configuration
+  servers_includelist: ["local"]  # Optional: Whitelist of MCP server names to allow
+  servers_excludelist: []  # Optional: Blacklist of MCP server names to exclude
+  
+  # Sub-agent configuration
+  subagent_paths: [".cecli/subagents"]  # Optional: Directories to search for sub-agent definitions
+  max_sub_agents: 3  # Optional: Maximum concurrent sub-agents (default: 3)
+  
   # Context blocks configuration
   include_context_blocks: ["todo_list", "git_status"]  # Optional: Context blocks to include
   exclude_context_blocks: ["symbol_outline", "directory_structure"]  # Optional: Context blocks to exclude
@@ -177,6 +186,10 @@ agent-config:
 - **`tools_includelist`**: Array of tool names to allow (only these tools will be available)
 - **`tools_excludelist`**: Array of tool names to exclude (these tools will be disabled)
 - **`tools_paths`**: Array of directories or Python files containing custom tools to load
+- **`servers_includelist`**: Array of MCP server names to allow (only these servers will be available)
+- **`servers_excludelist`**: Array of MCP server names to exclude (these servers will be disabled)
+- **`subagent_paths`**: Array of directories to search for sub-agent definition `.md` files
+- **`max_sub_agents`**: Maximum number of concurrent sub-agents (default: 3)
 - **`include_context_blocks`**: Array of context block names to include (overrides default set)
 - **`exclude_context_blocks`**: Array of context block names to exclude from default set
 
@@ -256,6 +269,7 @@ The following context blocks are available by default and can be customized usin
 - **`symbol_outline`**: Lists classes, functions, and methods in current context
 - **`todo_list`**: Shows the current todo list managed via `UpdateTodoList` tool
 - **`skills`**: Include skills content in the conversation
+- **`sub_agents`**: Include registered sub-agents in the conversation context
 
 When `include_context_blocks` is specified, only the listed blocks will be included. When `exclude_context_blocks` is specified, the listed blocks will be removed from the default set.
 
@@ -282,6 +296,14 @@ agent-config:
   tools_excludelist: ["command", "commandinteractive"]  # Optional: Blacklist of tools
   tools_paths: ["./custom-tools", "~/my-tools"]  # Optional: Directories or files containing custom tools
   
+  # Server configuration
+  servers_includelist: ["local"]  # Optional: Whitelist of MCP server names to allow
+  servers_excludelist: []  # Optional: Blacklist of MCP server names to exclude
+  
+  # Sub-agent configuration
+  subagent_paths: [".cecli/subagents"]  # Optional: Directories to search for sub-agent definitions
+  max_sub_agents: 3  # Optional: Maximum concurrent sub-agents (default: 3)
+  
   # Context blocks configuration
   include_context_blocks: ["todo_list", "git_status"]  # Optional: Context blocks to include
   exclude_context_blocks: ["symbol_outline", "directory_structure"]  # Optional: Context blocks to exclude

cecli/website/docs/config/subagents.md

@@ -21,7 +21,7 @@ Sub-agents can be used for:
 ┌─────────────────────────────────────────────────────────┐
 │                    TUI Session                           │
 │  ┌──────────────────────────────────────────────────┐   │
-│  │ SubAgentPills: ○ Primary  ● reviewer  ○ tester   │   │
+│  │ Mode: ○ primary  ◆ reviewer  ○ tester              │   │
 │  ├──────────────────────────────────────────────────┤   │
 │  │          OutputContainer (active agent)          │   │
 │  ├──────────────────────────────────────────────────┤   │
@@ -67,7 +67,7 @@ and suggestions for improvement.
 
 | Field | Required | Description |
 |-------|----------|-------------|
-| `name` | Yes | Unique name used to reference the sub-agent in commands and the Dispatch tool |
+| `name` | Yes | Unique name used to reference the sub-agent in commands and the Delegate tool |
 | `model` | No | Model override for this sub-agent. If omitted, inherits the parent agent's model |
 
 #### System Prompt
@@ -107,13 +107,13 @@ The most common way to use sub-agents. The primary agent waits for the sub-agent
 
 This sends the prompt to the reviewer sub-agent, which works autonomously and returns a summary when done.
 
-### Dispatching from the Primary Agent
+### Delegating from the Primary Agent
 
-The primary agent can also delegate work using the `Dispatch` tool. This enables the autonomous workflow:
+The primary agent can also delegate work using the `Delegate` tool. This enables the autonomous workflow:
 
 1. The primary agent analyzes a task
 2. It decomposes the work into sub-tasks
-3. It dispatches each sub-task to the appropriate sub-agent
+3. It delegates each sub-task to the appropriate sub-agent
 4. Sub-agents work independently and return their summaries
 5. The primary agent synthesizes the results
 
@@ -141,10 +141,10 @@ This is useful if a sub-agent is stuck, misbehaving, or you no longer need its w
 
 ### Switching Between Agents
 
-When sub-agents are active, the TUI shows a **SubAgentPills** bar at the top of the output area, displaying each agent as a clickable pill:
+When sub-agents are active, the TUI shows agent pills in the input container's border title, displaying each agent with status icons:
 
 ```
-┌─ [primary]  ● [reviewer]  ○ [tester] ──────────────────┐
+┌─ agent: ○ primary  ◆ reviewer  ○ tester ─────────────────┐
 ```
 
 - **Keyboard**: Use `Ctrl+Alt+Left` / `Ctrl+Alt+Right` to cycle through agents. Use `Ctrl+Alt+Up` to return to the primary agent.
@@ -157,13 +157,13 @@ Each agent gets its own output container. When you switch agents:
 1. The active container is shown; all others are hidden
 2. Your input is routed to the active agent
 3. Tool output, streaming responses, and task notifications are displayed in the correct container
-4. The SubAgentPills bar highlights the active agent
+4. Agent pills in the border title highlight the active agent
 
 ## Lifecycle and Limits
 
 ### Max Sub-Agents
 
-The `max_subagents` setting (default: 3) limits how many concurrent sub-agents can exist. This prevents resource exhaustion.
+The `max_sub_agents` setting (default: 3) limits how many concurrent sub-agents can exist. This prevents resource exhaustion.
 
 When the limit is reached:
 
@@ -178,7 +178,7 @@ When the limit is reached:
 
 ## Restrictions
 
-- **No nested sub-agents**: Sub-agents cannot spawn further sub-agents. The `Dispatch` tool is excluded from sub-agent tool schemas.
+- **No nested sub-agents**: Sub-agents cannot spawn further sub-agents. The `Delegate` tool is excluded from sub-agent tool schemas.
 - **TUI-dependent**: Sub-agent container switching and the reap command depend on the TUI. Running in headless or non-TUI modes may not support these features.
 
 ## Examples
@@ -222,8 +222,8 @@ happy paths. Use the project's existing testing patterns and conventions.
 
 By defining multiple sub-agents, you can get different perspectives on the same code:
 
-1. Dispatch a **reviewer** to analyze security concerns
-2. Dispatch a **tester** to identify test gaps
+1. Delegate to a **reviewer** to analyze security concerns
+2. Delegate to a **tester** to identify test gaps
 3. The primary agent synthesizes both reports into an action plan
 
 ## See Also