Research
A shared Slack workspace for collaborating agents
Slack channels, messages, and agent descriptions form a communication layer for human–agent and agent–agent work.
SlackAgents explores a familiar setting for multi-agent systems: the workspace where people already exchange requests, share context, and coordinate tasks. The framework places configurable AI agents inside Slack and uses its communication structures to connect their work.
An agent joins the conversation
Each agent has an identity and a configuration describing its capabilities. A Slack application connects that agent to workspace messages. Users interact through the same channels and conversations that carry the rest of their work.
The open implementation includes agent configuration, Slack application setup, custom tools, and examples. Those pieces make the connection between a model-backed specialist and its messaging environment available for inspection and extension.
Colleagues become part of the configuration
Multi-agent collaboration uses a set of colleagues identified by their workspace user IDs, names, and descriptions. This gives an agent information about the people and other agents participating in a channel. A task can draw on several specialists whose roles are described within that shared environment.
The communication layer becomes an architectural choice. Messages carry requests and responses, while the channel supplies a common place for participants to follow the work. Human contributions and agent contributions enter the same interaction.
Building on an existing work surface
The research contributes a library and working demonstrations for developing, deploying, and managing agents through Slack. Its examples provide a starting point for exploring tool-backed assistants and collaborative workflows.
For a developer, this approach makes several design questions tangible: how to describe each specialist, which messages should reach it, and how the next participant receives a result. Keeping those choices explicit helps explain a collaboration as a sequence of concrete interactions.
SlackAgents was developed at Salesforce AI Research by a team including Weiran Yao and Frank Wang and presented in the EMNLP 2025 System Demonstrations track. Read the paper and explore the library.