Feature Request: "AutoClaw Mode" — Native Hermes & OpenClaw Agent Integration Inside ZCode IDE

Summary

Introduce an "AutoClaw Mode" into ZCode IDE that natively integrates Hermes Agent and OpenClaw agents directly into the development environment. Rather than treating these agents as external services accessed through a browser, AutoClaw Mode would embed them as first-class IDE citizens — giving developers the full power of AI agent-driven coding, workspace management, and autonomous task execution without ever leaving ZCode.

---

Problem Statement

Today, users who want to leverage Hermes Agent, OpenClaw, or AutoClaw's agent ecosystem must context-switch between two separate environments:

• ZCode IDE — for writing, editing, building, and debugging code locally.
• AutoClaw Web / External Agent Interface — for deploying agents, running automated workflows, managing agent clones, and orchestrating multi-agent tasks.

This split creates significant friction:

1. Context fragmentation. A developer working in ZCode must manually transfer context (file paths, code snippets, task descriptions, project state) to the external agent interface. This is error-prone and breaks flow state.
2. No real-time IDE awareness. External agents lack deep integration with the local project — they don't automatically see file changes, terminal output, git status, or the live state of the codebase inside ZCode.
3. Redundant tooling. ZCode already has a powerful agent system (subagent dispatch, tool calling, MCP servers). Rebuilding similar capabilities in an external interface is duplicative when the two could share a single unified pipeline.
4. Discovery gap. Many ZCode users may not even know Hermes and OpenClaw exist, or may not understand how to connect them to their ZCode workflow. Native integration eliminates the discovery and setup barrier entirely.

---

Proposed Feature: AutoClaw Mode

What It Is

AutoClaw Mode is a toggleable workspace mode inside ZCode (similar to the previously proposed "Office Mode") that activates the Hermes/OpenClaw agent layer on top of ZCode's existing tooling. When enabled:

• The IDE gains access to specialized agent profiles optimized for different coding and automation tasks.
• Agents operate with full project context — they can read, write, and modify files in the active workspace, interact with the terminal, run tests, and report results back directly in the IDE.
• Multi-agent workflows (e.g., one agent writes code, another reviews it, a third runs tests) can be orchestrated from within ZCode itself.

Core Capabilities

1. Native Agent Panel

A dedicated sidebar or panel inside ZCode that provides:

• Agent selection: Choose from available agent profiles (Hermes, OpenClaw, custom clones) directly from the IDE.
• Agent status: Real-time visibility into which agents are running, their current task, progress, and output — all within the IDE UI.
• One-click dispatch: Send tasks to agents via a natural-language input bar, file context menu (right-click → "Ask Agent"), or inline code selection (highlight code → "Refactor with Hermes").

2. Full Workspace Awareness

Agents in AutoClaw Mode operate with deep project context:

• Automatic access to the active project's file tree, open tabs, and recent edits.
• Ability to read and write files directly in the workspace (no manual file path passing).
• Access to terminal output, git diffs, build logs, and test results.
• Awareness of MCP server connections configured in ZCode — agents can leverage the same tools the parent IDE session has.

3. Multi-Agent Orchestration

Enable agent-to-agent collaboration within ZCode:

• Sequential pipelines: Agent A generates code → Agent B reviews → Agent C runs tests. The full pipeline runs inside ZCode with results surfaced at each stage.
• Parallel dispatch: Multiple agents work on independent tasks simultaneously (e.g., one refactors module A while another writes tests for module B), with results merged back into the workspace.
• Hierarchical delegation: A parent agent (Hermes) can spawn child agents (OpenClaw instances) for subtasks, all managed from the IDE's agent panel.

4. Agent Clone Management

For users who have set up agent clones in AutoClaw (personalized agent configurations with custom instructions, tool preferences, and workflow patterns), AutoClaw Mode would allow:

• Importing and managing clones directly from ZCode.
• Assigning specific clones to specific tasks or file types (e.g., "Always use my Python Expert clone for .py files").
• Syncing clone configurations between AutoClaw Web and ZCode so changes are reflected everywhere.

5. Persistent Agent Sessions

Unlike the current subagent model where agents are fire-and-forget, AutoClaw Mode would support:

• Long-lived agent sessions that persist across multiple tasks within a single ZCode session.
• Conversation memory — an agent remembers the context of prior tasks in the same session, reducing redundant explanations.
• Background agents that continue working while the developer focuses on other files or tasks in the IDE.

---

How It Compares to Today's Workflow

Aspect	Current (ZCode + External AutoClaw)	Proposed (AutoClaw Mode)
Agent access	Separate browser tab / web interface	Integrated sidebar panel in IDE
Project context	Manual copy-paste of files, paths, descriptions	Automatic — agents see the full workspace
File editing	Agent outputs code snippets to copy back	Agent edits files directly in the project
Terminal / tools	No access to ZCode terminal or MCP servers	Full access — terminal, git, MCP, build tools
Multi-agent workflows	Manual coordination across tabs	Orchestrated pipelines from a single UI
Session persistence	Separate session management	Unified session within ZCode
Discovery	User must know about and navigate to AutoClaw	Available as a built-in mode — zero setup

---

Why This Matters

1. Unified developer experience. The best developer tools are the ones that keep you in flow. Forcing users to leave their IDE to interact with agents is a flow-breaking context switch that adds friction to every agent task. Bringing agents into the IDE eliminates this entirely.

2. Leverage existing ZCode infrastructure. ZCode already has file management, terminal access, MCP server integration, tool calling, and a subagent system. AutoClaw Mode layers the Hermes/OpenClaw agent intelligence on top of infrastructure that already exists — minimal new plumbing, maximum new capability.

3. Competitive positioning. Claude Code has agents in the terminal. Cursor has AI inline. No one has a full multi-agent orchestration system native to the IDE. AutoClaw Mode would be a unique differentiator that no other coding tool currently offers.

4. Bridge for non-AutoClaw users. ZCode users who have never used AutoClaw would discover Hermes and OpenClaw organically — just by toggling a mode. This drives adoption of the entire Z.ai ecosystem from within the most-used entry point.

---

Suggested Implementation Approach

Phase	Scope	Notes
Phase 1	Agent panel + single-agent dispatch (Hermes or OpenClaw) with workspace file access	MVP — prove the integration works inside ZCode
Phase 2	Multi-agent orchestration, persistent sessions, terminal/MCP access	Full power user experience
Phase 3	Agent clone sync, background agents, cross-device session continuity	Advanced features

---

Environment

ZCode 3.4.2.3624 — Feature applicable to all platforms.

---

Regards,
Roman
http://www.rommark.dev

VIbe Coders Telegram Channel: https://t.me/VibeCodePrompterSystem

No files, tests, or implementation entry points are named. Start by reviewing ZCode's existing subagent dispatch, tool-calling, MCP, file-management, and terminal systems, then narrow the request to a defined Phase 1; done should be a scoped, testable single-agent integration with workspace access.