Feature Suggestion: LLM Composition and Orchestration Framework within Kilo Code

[riccardomorabito]

As LLMs grow more specialized, many developers are leveraging multiple models in tandem—using each for its strengths. It would be extremely powerful if Kilo Code provided native support for composing and orchestrating different LLMs (e.g., Claude 4 Sonnet, Mistral) during software development workflows.

Developers could create pipelines or assign models to specific tasks in Kilo Code—e.g., using Claude for system-level design and GPT for implementation details.

Multi-LLM Routing Layer: A config layer where developers can define which model handles what type of request.

Prompt Routing DSL or UI: A minimal language or UI element to define orchestration logic (e.g., "Claude for codegen, GPT for planning").

Session Memory Sync: Shared memory or context passing between models when chaining requests (e.g., plan → scaffold → implement).

Claude 4 Sonnet Integration: Native support or API bridge to integrate Claude 4 Sonnet into the Kilo Code toolchain.

🔍 Benefits

Maximizes strengths of each model: Developers aren’t limited to a one-size-fits-all LLM.

Boosts development speed and flexibility.

Encourages modular prompt design and more sophisticated workflows.

Future-proofing for when even more models become part of the toolset.

Would love to hear what the team and community think!