agent-factory-discord
v1.0.0
Published
Multi-agent Discord bot with debate, routing, and local/cloud agent support
Readme
agent-factory-discord
Multi-agent Discord bot for routing prompts across local and cloud agents, with optional debate workflows and review steps.
This package is a good fit for teams that want one Discord interface over several model backends instead of binding themselves to a single provider.
Features
- Auto-routing by task type
- Explicit agent selection with
!agent:commands - Debate mode with multiple agents
- Optional red-team review flow
- Support for local models, remote hosted models, CLI-based agents, and API-based agents
Supported Agent Types
- local Ollama models
- remote HTTP-backed local models
- Claude-backed flows
- Codex-backed flows
- Gemini-backed flows
The package is designed so you can enable only the agents you actually use.
Example Commands
What is X?
!claude: explain this stack trace
!local: summarize this README
!gemini: analyze this long PDF
!debate: should we ship this API design?
!debate claude vs local: compare these approaches
!debate claude vs local 3: evaluate this refactor planSetup
- Copy
.env.exampleto.env - Create a Discord bot and add it to your server
- Configure the agents you want to enable
- Start the bot
cp .env.example .env
npm install
npm run devRouting Model
The router is intended to favor cheap or local execution first for lightweight tasks, then escalate to stronger models for long documents, vision, or higher-context work.
Environment
Common settings include:
DISCORD_TOKENDISCORD_CLIENT_IDDISCORD_GUILD_IDDISCORD_TASK_CHANNEL_IDAGENT_ROUTER_ENABLEDOLLAMA_URLGEMINI_API_KEY
Some optional settings in .env.example refer to a remote model endpoint under the label BEAST_*. You can treat that as a generic remote local-model service and rename those environment variables in your own deployment if you prefer.
Notes
- This package assumes Discord as the interaction surface
- It is best used as an app starter or template rather than a drop-in library
- You should review agent permissions carefully before enabling execution-oriented flows
