Ollama's OpenCode now supports subagents—letting you parallelize complex tasks like research, refactoring, and code reviews by spinning up multiple AI agents that work simultaneously.
What It Is
OpenCode in Ollama now supports running subagents, which are essentially child AI processes that can tackle different parts of a larger task in parallel. Instead of one AI working through a long context window sequentially, you can break complex work into chunks—research, refactoring, code review—and have specialized agents tackle each piece simultaneously, then combine results.
How This Helps Today
For developers, this means faster code reviews on large pull requests—one agent checks for security issues, another for performance, another for style. Research tasks that would take hours can be parallelized across multiple subagents gathering different sources. Refactoring large codebases becomes more manageable when different agents handle different modules. The key benefit: complex tasks that exceed context limits or would take too long sequentially can now be distributed.
The Context
This addresses one of the biggest limitations of current AI coding tools—context window constraints and sequential processing. By enabling parallel subagents, Ollama is essentially bringing distributed computing concepts to AI-assisted development. This positions it against tools like GitHub Copilot Workspace and Cursor, but with the advantage of running locally and supporting parallel execution patterns.
What to Watch
Coordination overhead—do subagents stay coherent or step on each other's work? Resource usage—how much RAM/CPU does running multiple local LLMs consume? Result quality—does parallelization improve or degrade output compared to single-agent sequential processing? Integration with existing workflows—can this plug into CI/CD or does it require new tooling? And whether the speed gains justify the complexity for typical development tasks.