Multiple agents during operation

[codecanyon128]

Describe the feature or problem you'd like to solve

No response

Proposed solution

No response

Example prompts or workflows

We want a new feature that lets you choose a specific modal that does a particular thing, like Gemina 3 for the front-end and Opus for the back-end. I think it's a fantastic feature.

Additional context

No response

[coryzibell]

+1 on this. The use case is cost/latency optimization - not every task needs the most capable model.

Proposed approach: YAML frontmatter in agent definitions

Claude Code solves this with a model field in their agent markdown files:

---
name: fast-linter
description: Quick syntax checks
tools: Bash, Read
model: gpt-4o-mini
---

You are a fast linter for quick syntax validation...

And for heavier work:

---
name: architect
description: System design decisions
tools: Read, Glob, Grep
model: o1
---

You are a senior architect making structural decisions...

Their docs on this:

• Slash commands - model frontmatter
• Subagents - model selection

Implementation options:

model value	Behavior
Model ID (e.g., gpt-4o-mini)	Use that specific model
inherit	Use the session's current model
Omitted	Default model for agents

This would let users optimize their agent fleet - fast/cheap models for atomic tasks, capable models for complex reasoning. The AGENTS.md format already supports YAML frontmatter, so this could slot in naturally.

[examon]

This is now supported as of v0.0.415. Custom agents can specify a model: field in their YAML frontmatter to choose which model that agent should use, making it possible to use different models for different agents in the same setup. If you believe this isn't resolved, please feel free to reopen this issue.