Ollama v0.32.11 brings DeepSeek Harness and Muse Code agent support
The local LLM runner now supports agentic coding from DeepSeek and Meta, plus web search.
Ollama, the widely adopted local LLM runner with over 179k GitHub stars, has shipped v0.32.11. The release integrates two major agentic coding frameworks: DeepSeek Harness, DeepSeek's open-source agent harness, and Muse Code, Meta's agentic coding CLI. This means developers can now pair locally hosted models with purpose-built agents that autonomously plan and execute code tasks—right from the command line.
Beyond agent support, Ollama upgraded its OpenAI-compatible Responses API to support web search, giving local applications access to live information without leaving the OpenAI protocol. The release also includes updates to Muse Glimmer templates, which refine how Muse Code renders and interacts with code. For teams exploring AI-assisted development, v0.32.11 turns Ollama into a more complete platform for running agentic workflows locally, reducing reliance on cloud services and keeping data on-premises.
- Adds support for DeepSeek Harness, DeepSeek's open-source agent harness
- Integrates Meta's Muse Code agentic coding CLI for local coding agents
- OpenAI-compatible Responses API now supports web search; Muse Glimmer templates updated
Why It Matters
Brings production-grade agentic coding to local LLM setups, expanding Ollama's role beyond simple model serving.