Server facts
context-rot-detection MCP Server
Project
- Author
- milos-product-maker
- License
- Not specified
- Category
- AI Models & Agents
Tech
- Runtime / language
- Not specified
- Transports
- stdio
- Remote available
- No
- Authentication
- Not specified
Activity
- GitHub stars
- 13
- Last commit
- Data as of
Tags
- agent health
- ai agents
- context engineering
- context rot
- context window
- llm
- mcp
- model context protocol
Overview
Long agent sessions can accumulate context that becomes harder to use effectively. Context Rot Detection exposes check_my_health to assess that condition from an estimated token count, with optional details about the model, elapsed session time, tool calls, task context, and agent identity. Its output combines a health score with utilization metrics, estimated retrieval degradation, risk associated with information in the middle of the context, and fatigue indicators. Recommendations can suggest summarizing older material, moving information to external memory, creating checkpoints, or dividing work into smaller tasks; these are suggestions rather than recovery actions performed by the service.
Model handling includes curated degradation profiles and context-window lookup from HuggingFace repository configurations, with SQLite caching and conservative defaults when resolution fails. For ongoing monitoring, get_health_history retrieves an identified agent’s earlier checks and aggregate statistics. Operators can use get_service_stats to inspect call totals, agent counts, average scores, model and status distributions, and recent activity. The service communicates over stdio, while health-history storage can use a configured SQLite database or remain ephemeral in memory.
context-rot-detection setup notes
- context-rot-detection records stdio only; confirm a network transport upstream before planning remote access.
- context-rot-detection has no recorded license; confirm reuse terms in its repository.