dsh-context is an open-source plugin for DeepSeek Harness that provides detailed visualization and management of AI agent context usage. It helps developers understand how conversation history, tool responses, files, and system instructions contribute to an agent's context window. A dedicated dashboard displays token consumption, context composition, trends, and significant events throughout a session. Cross-session analytics reveal usage patterns, estimated costs, cache performance, and activity over time. The plugin also visualizes relationships between parent agents and subagents in multi-agent workflows. Users can inspect context changes through an integrated tab, sidebar panel, or slash command. Its browser-based interface provides tools for investigating context growth, compression, and session behavior without modifying the underlying AI models.
Features
- AI Agent Context Usage Dashboard
- Token Consumption and Cost Analytics
- Context Composition and Evolution Tracking
- Cross-Session Activity Visualization
- Multi-Agent Hierarchy and Fleet Monitoring
- Integrated Context Panel and Slash Commands