Enable workspace-agent collaboration, Hosted MCP, and Mem0 in Qwen Code v0.25.0
Qwen Code v0.25.0 (no known breaking changes) adds local workspace-agent collaboration, A2A sharing, bundled Mem0, private Hosted MCP, Hosted tool approvals in WebShell, and default durable local-process recovery. Adopt these so agents share workspaces, recall context, and survive reboots.Why this loop
v0.25.0 is a managed-agent and workspace-collaboration jump, not a breaking upgrade. Local workspace-agent collaboration and A2A sharing let multiple agents work in one workspace and share access; session creators can submit, cancel, and rename workspace-bound sessions. Mem0 now ships in the main CLI, with an opt-in selector skip on a unique recall hit, and the prompt path answers from conversation history before investigating—so stop restarting context every session. Durable local-process and trusted reboot recovery default on, with verified W1a cold restore, offline W1b bundles, hosted file history/undo, durable remote Shell results, and hosted hooks. Gate tools with private Hosted MCP (H1), Hosted tool approvals (D6a) answered in the Managed panel, durable permission Actions (D6b), persisted workspace tool profiles, and Runtime-worker tool execution (M5a). Code Mode now allows Freeform input and concurrent Bash. Inspect runs with the collapsible trajectory waterfall and projected durable tool results in WebShell. Daemon extension settings are isolated per hosted workspace so .env no longer leaks. Do not invent behavior from the 22 omitted changelog entries.
Proposed actions
- Upgrade Qwen Code to v0.25.0 (no known breaking changes) and turn on local workspace-agent collaboration so multiple agents share one workspace instead of isolated sessions.
- Enable A2A access and sharing for workspace agents, then let each Workspace-bound Session's creator submit, cancel, and rename those sessions with teammates.
- Use Mem0 now bundled in the main CLI; enable the opt-in selector skip for a unique new recall hit, and instruct the agent to answer from conversation history before investigating.
- Enable the private Hosted MCP runtime (H1), require Hosted tool approvals (D6a) answered in the WebShell Managed panel, persist Workspace session tool profiles, and run Managed session tools in a Runtime worker (M5a).
- Keep durable local-process and trusted reboot recovery on (now default); after a crash use verified W1a workspace cold restore, offline W1b recovery bundles, hosted file history/undo, and durable remote Shell result delivery.
Agent prompt
Paste into your agent or query via MCP (get_agent_prompt) — free, no extra AI costCopilot Chat / Agent mode
DevAgentRadar → GitHub Copilot
You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.
Context
Assistant: Qwen Code Proposal: Enable workspace-agent collaboration, Hosted MCP, and Mem0 in Qwen Code v0.25.0 Summary: Qwen Code v0.25.0 (no known breaking changes) adds local workspace-agent collaboration, A2A sharing, bundled Mem0, private Hosted MCP, Hosted tool approvals in WebShell, and default durable local-process recovery. Adopt these so agents share workspaces, recall context, and survive reboots. Primary source: https://github.com/QwenLM/qwen-code/releases/tag/v0.25.0
Why it matters
v0.25.0 is a managed-agent and workspace-collaboration jump, not a breaking upgrade. Local workspace-agent collaboration and A2A sharing let multiple agents work in one workspace and share access; session creators can submit, cancel, and rename workspace-bound sessions. Mem0 now ships in the main CLI, with an opt-in selector skip on a unique recall hit, and the prompt path answers from conversation history before investigating—so stop restarting context every session. Durable local-process and trusted reboot recovery default on, with verified W1a cold restore, offline W1b bundles, hosted file history/undo, durable remote Shell results, and hosted hooks. Gate tools with private Hosted MCP (H1), Hosted tool approvals (D6a) answered in the Managed panel, durable permission Actions (D6b), persisted workspace tool profiles, and Runtime-worker tool execution (M5a). Code Mode now allows Freeform input and concurrent Bash. Inspect runs with the collapsible trajectory waterfall and projected durable tool results in WebShell. Daemon extension settings are isolated per hosted workspace so .env no longer leaks. Do not invent behavior from the 22 omitted changelog entries.
Suggested actions
- Upgrade Qwen Code to v0.25.0 (no known breaking changes) and turn on local workspace-agent collaboration so multiple agents share one workspace instead of isolated sessions.
- Enable A2A access and sharing for workspace agents, then let each Workspace-bound Session's creator submit, cancel, and rename those sessions with teammates.
- Use Mem0 now bundled in the main CLI; enable the opt-in selector skip for a unique new recall hit, and instruct the agent to answer from conversation history before investigating.
- Enable the private Hosted MCP runtime (H1), require Hosted tool approvals (D6a) answered in the WebShell Managed panel, persist Workspace session tool profiles, and run Managed session tools in a Runtime worker (M5a).
- Keep durable local-process and trusted reboot recovery on (now default); after a crash use verified W1a workspace cold restore, offline W1b recovery bundles, hosted file history/undo, and durable remote Shell result delivery.
After you finish
Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.
Tell the human: open https://devagentradar.com/proposals/qwen-code-v0-25-0-enable-workspace-agent-collaboration-hosted-mcp-and-me and mark Applied, Skipped, or Failed. Proposal id: 51a69c6e-7503-4f4e-9b68-d747a830a3d9
Your job
- Restate the change in one sentence.
- Propose a minimal plan for my repo (or a throwaway pilot).
- Implement only what I approve; prefer small diffs and tests.
- Call out risks (permissions, breaking APIs, cost).
Start by confirming you understood the proposal.
Your loop
This browser · no sign-in · not shared as “you”After you run the prompt
Only you can mark this. Agents cannot write your loop.
Your decision stays on this device. A public tally appears after a few votes.