contextenginelisted

# ContextEngine MCP Local-first context layer for AI agents. Use it when work needs shared project memory, append-only captures, reviewable patch proposals, team session learnings, or legacy `agent-loop-mcp` compatibility. ## When to use - The user references an ongoing project and you need shared context across sessions. - You need to capture notes, decisions, or outcomes in a structured project memory. - You want to propose broader context updates without applying them blindly. - A team project uses multiple agents or multiple teammate machines and needs shared learnings from solved or failed work. - You are continuing an older workflow that still depends on the `agent-loop-mcp` session tools. ## Context tools | Tool | Use it when | |------|-------------| | `init_context` | Starting a new project memory. | | `read_context` | Beginning a session or checking a specific topic. | | `append_capture` | Saving a raw note, transcript, or user capture. | | `search_context_topics` | Looking up earlier notes, goals, or decisions. | | `log_agent_outcome` | Recording a finished subtask or result under a topic. | | `compact_topic` | Replacing a long section with a curated summary. | | `propose_context_patch` | Suggesting a broader rewrite to `context.md` for approval. | | `list_pending_patches` | Showing what is waiting for review. | | `reject_context_patch` | Discarding a pending proposal. | | `apply_context_patch` | Applying an explicitly approved proposal. | | `undo_context_patch