Coke / OpenCoke
Coke is an AI Agent system for everyday IM scenarios, built to help AI understand context, remember tasks, proactively follow up, call tools, and collaborate with users through real communication channels.
Why it matters
Beyond chatbot replies
The product surface is IM, but the work is systems.
Multi-channel IM access
Context-aware conversation
Proactive reminders
Tool and skill calls
Memory and context
Deployment validation
A concise runtime boundary map.
Gateway / Bridge
Normalizes messages from IM channels and shared-channel integrations.
Agent System
Routes intent, retrieves context, generates replies, and coordinates agent workflows.
Reminder System
Owns reminder state, scheduling, firing, and proactive follow-up boundaries.
Tool / Skill System
Loads reusable capabilities and connects agent reasoning with real operations.
Context / Memory
Keeps task, user, relationship, channel, and time context inspectable.
From request to follow-up.
A user can create a reminder in an IM channel. Coke interprets the task, stores the schedule, keeps the relevant context, and sends the follow-up message through the channel path.
1. User intent
A natural-language request enters through an IM channel.
2. Agent routing
The agent identifies whether the message is conversation, task, reminder, or tool work.
3. Runtime state
Reminder and context systems persist the task boundary.
4. Follow-up
The system sends a proactive message when the condition is met.
Active development.
Coke is under active development. The current focus is reliability, reminder behavior, multi-channel deployment, public demos, and open-source documentation.