Getting multiple AI coding agents to work together usually means stitching together MCP servers, custom APIs, or shared file systems with careful orchestration logic. A simpler pattern is gaining traction among developers: just use Redis.
The approach is almost comically straightforward. You spin up a Redis instance, hand the connection string to your AI agent, and tell it to treat Redis as a shared message board. One agent becomes the leader, posting task assignments to Redis keys. Worker agents poll for new tasks, claim them, and write results back. No custom framework, no agent-to-agent protocol, no dependency graph to maintain.
The trick that makes this work is that modern AI coding agents like Cursor are already good at following instructions about how to communicate. You give the leader agent a prompt describing the coordination protocol - how workers announce themselves, where tasks get posted, how to signal completion - and the agent generates the actual Redis commands on the fly. The "framework" is just a prompt and a Redis URL.
Where This Actually Helps
The sweet spot is projects where you want multiple agents working on different parts of a codebase simultaneously. One agent handles frontend components while another writes backend endpoints, with Redis acting as the shared brain for who's doing what. Compared to building a proper orchestration layer with something like LangGraph or CrewAI, the Redis approach trades sophistication for speed of setup. You can have agents coordinating in minutes instead of days.
The obvious limitation: this is coordination, not collaboration. Agents can divide work and avoid stepping on each other, but they're not reasoning together about architecture decisions or resolving merge conflicts. For that, you still need human oversight or a more structured framework.
For solo developers or small teams already running Redis, this is a zero-cost experiment worth trying. The pattern works with any AI coding tool that can execute shell commands or call APIs - Cursor, Claude Code, Aider, or Cody could all participate in the same Redis-coordinated workflow.