Description
Currently, openab is configured with a single [agent] command, which creates a 1:1 mapping between an openab instance and an ACP-compatible CLI (e.g., Gemini CLI).
For users who maintain multiple specialized agents (e.g., Gemini for general tasks, Codex for legacy code, and Kiro for system-level tasks), running multiple openab instances is resource-intensive and complex to manage.
Proposed Feature
I would like to propose a feature that allows:
- Defining multiple agents in the configuration (e.g., using
[[agents]] or a map of agents).
- Dynamic Routing of requests to specific agents based on:
- Mentions: e.g.,
@Gemini help me vs @Kiro fix this.
- Channel Mapping: Mapping specific Discord/Slack channels to specific agents (e.g., Channel A -> Agent 1, Channel B -> Agent 2).
- Default Agent: Specifying a default agent with fallback mechanisms.
This would make openab a much more powerful "multiplexer" or "bridge" for the ACP ecosystem, allowing users to consolidate their infrastructure while leveraging the strengths of different agents.
Thank you for the great work on this project!
Description
Currently,
openabis configured with a single[agent]command, which creates a 1:1 mapping between anopenabinstance and an ACP-compatible CLI (e.g., Gemini CLI).For users who maintain multiple specialized agents (e.g., Gemini for general tasks, Codex for legacy code, and Kiro for system-level tasks), running multiple
openabinstances is resource-intensive and complex to manage.Proposed Feature
I would like to propose a feature that allows:
[[agents]]or a map of agents).@Gemini help mevs@Kiro fix this.This would make
openaba much more powerful "multiplexer" or "bridge" for the ACP ecosystem, allowing users to consolidate their infrastructure while leveraging the strengths of different agents.Thank you for the great work on this project!