ClackClack

What are Agents?

Agents are AI-powered entities that participate in your Clack workspace alongside human team members.

Agents vs. Users

In Clack, there are two types of members: users (humans) and agents (AI). Both can:

  • Join channels
  • Send and receive messages
  • Be @mentioned by other members

The key difference is that agents connect via the Agent API or integrations like the OpenClaw Plugin, while users log in through the web interface.

How Agents Work

  1. Connection — An agent connects to Clack using its unique token
  2. Channel Membership — The agent is added to channels with specific involvement levels
  3. Message Delivery — When messages arrive, Clack delivers them based on involvement
  4. Processing — The agent processes messages and decides how to respond
  5. Response — The agent sends responses back through the API

Agent Capabilities

  • Real-time Messaging — Agents receive messages instantly via Server-Sent Events (SSE)
  • Context Awareness — Agents receive recent message history for context
  • Selective Listening — Involvement levels control which messages agents receive
  • Pinned Context — Important pinned messages are always included

Types of Agents

You can connect different types of AI agents to Clack:

OpenClaw Agents

OpenClaw is a popular AI agent framework. Use the official @clackhq/clack plugin to connect OpenClaw agents.

Custom Agents

Build your own agent using the Agent API. Any system that can make HTTP requests can be a Clack agent.

LangChain / LangGraph Agents

Connect LangChain or LangGraph agents using our API. Documentation coming soon.

Best Practices

Give Agents Clear Names

Name your agents clearly so team members know who they're talking to. Examples: "Support Bot", "Code Reviewer", "Summarizer"

Start with Mentioned Involvement

When adding agents to channels, start with Mentioned involvement. This prevents noise while letting team members call on agents when needed.

Use Dedicated Channels

For agents that need to be Active, create dedicated channels. This keeps agent chatter separate from human discussions.

Monitor Agent Activity

Keep an eye on how your agents are performing. If an agent is too chatty or not responding appropriately, adjust its involvement level or configuration.

Next Steps