Feature Request: Allow Changing the Coding Agent Profile After Task Creation

[Lobarr]

Summary

I would like to request the ability to change the coding agent's profile on a task after it has been created. The current behavior locks the profile to the one selected at the time of task creation. This prevents users from switching models for reasons of task evolution, API limits, or cost.

---

Motivation

As a task progresses, the needs can change, but so can the external circumstances. The ability to switch agent profiles is crucial for two primary reasons:

1. Workflow Flexibility: A task might begin with a high-level "planning" agent and later require a specialized "debugging" or "code-refactoring" agent. Forcing users to create a new task just to switch agents leads to redundant task cards, context loss, and a cluttered board.

2. Economic and Practical Necessity: AI models and their APIs are not free. Users often have credit or API limits on different platforms (e.g., OpenAI, Anthropic, Google Gemini). It is common to hit an API limit on one platform while still having credit on another. The current static profile locks a user into a single model, forcing them to either wait for their API limits to reset or manually recreate the task on a different platform. This is a significant source of friction.

Allowing a change of agent profile mid-task would provide a seamless solution, enabling a more dynamic and flexible workflow that accounts for both the evolving nature of development tasks and the practicalities of managing different AI service accounts.

---

Proposed Functionality

The goal is to make the PROFILE field editable on existing tasks, providing a simple way to switch models.

1. Editable Profile Field: The PROFILE field in the task details view should become a clickable or editable element.
2. UI Element: This could be a dropdown menu that appears when the user clicks on the current profile name (qwen-code in the screenshot), or a small "edit" icon next to it.
3. Profile Selector: Clicking the edit element would display a list of all available agent profiles and models.
4. Preserve History: The most critical part of this feature is that changing the profile should not erase any existing chat history or code. The new agent should be able to see the full context of the previous agent's work. The system should simply add a note to the chat log, such as "Agent profile switched to 'Bug Fixer' at 8:41 PM."
5. Seamless Handoff: The new agent would then take over, leveraging the conversation history to continue the task from where the previous agent left off.

---

Benefits

• Enhanced Flexibility: Easily switch between specialized agents or models as needed.
• Improved Context Continuity: Agents can build upon each other's work, leading to more efficient and accurate results.
• Cost and API Management: Users can switch to a different model or provider when they hit an API limit or simply want to use a less expensive model for a particular sub-task.
• Streamlined Workflow: Reduces the need to create new tasks, keeping the board clean and focused.

[LSRCT]

We support different variants for profiles which can be changed after task creation, we ship claude code with a Default and a plan variant, these can be switched between at any point. You could create a debugging variant in the same way.

Note that these variants need to use the same cli, you can have multiple variants for e.g. gemini cli and switch between them, you can't go from gemini cli to claude code currently.

[Lobarr]

thanks for the quick response! oh i didn't realize that, it's great variants can be change. i should clarify that what you said we can't do today is what i'm requesting here

[ggordonhall]

If you want to plan with one agent and implement with another you could give the VK MCP a go. Make sure that it is installed (in the MCP Servers tab) then, when you're happy with your plan, instruct the agent to create a ticket. You should see one appear on the VK board and you can you any agent you want to implement it.

[Lobarr]

that still doesn't solve the problem i'm describing here iiuc

here's something i experienced which inspired this:

i was in the middle of a coding session with claude code and all of a sudden, my pro plan got rate limited for a few hours.

with the way things work today, i'll be stuck on that task (the specific snapshot claude code was working on) until my rate limit gets lifted.

or is there a way around this?

[leewardbound]

Im having the same problem as @Lobarr above - I am hitting "5 hour limit reached" very quickly (in less than 2-3 chats of moderate/reasonable length), despite not really hitting my token limits, and what I want to do here is switch to another model-or-agent mid-chat. Currently I am just unable to work on this task at all (in vibekanban) for the rest of the day.

[RIP21]

This would be very useful for just continuing work on the existing task after rate limiting using different agent like gemini instead of claude.

Also, would be awesome to gracefully retry with different agent, or let it wait till the timer resets and start again later automatically.

[stunningpixels]

We'll be allowing multiple coding agent sessions (and from different coding agents) on the same worktree in the future (4-8 weeks estimate)