Memory + context infrastructure for AI agents. LLM Wiki auto-sync, 15 content-aware summarizers, hybrid search (BM25 + vector + optional LLM judge)
Context Mem MCP server is a locally run integration for AI assistants that speak the Model Context Protocol. Memory + context infrastructure for AI agents. LLM Wiki auto-sync, 15 content-aware summarizers, hybrid search (BM25 + vector + optional LLM judge), entity intelligence, decision trails, Obsidian plugin. Fully local, zero cost. MIT.
Two problems with today's AI tooling that no one has solved together in a single package.
Once Context Mem is connected, these are the calls the assistant has available:
Benchmark — Retrieval RecallLongMemEval — 97.8% R@5LoCoMo — 98.1% R@10MemBench — 98.0% R@5ConvoMem — 97.7% R@10observe — Store observation with auto-summarization, importance scoring, entity extraction, topic detectionrecall — Retrieve verbatim content by filter (importance, type, flag, time)search — Hybrid search (BM25 + vector + optional LLM judge)ask — Natural language Q&A over the full memory storetimeline — Reverse-chronological observations with importance badges and flagsstats — Token economics for current session (raw vs. compressed)save_knowledge — Save a knowledge entry with contradiction detection + temporal validity windowsInstallation goes through your MCP client rather than a global install: point it at context-mem on npm and it is fetched when the client starts. The copy-paste blocks for Claude Desktop, Claude Code and Cursor are further down this page.
This sits in the knowledge and memory group, where several servers overlap in what they claim to do but differ sharply once you actually set them up. Context Mem's toolset — Benchmark, LongMemEval, LoCoMo and 11 more — is a fair guide to whether it matches your workflow. It is maintained by jubakitiashvili; worth a glance at recent repository activity before you build anything load-bearing on it.
This entry was verified against Context Mem's own documentation before publication; SyncDev keeps the directory reviewed rather than auto-generated.
| Tool | What it does |
|---|---|
| Benchmark | Retrieval Recall |
| LongMemEval | **97.8% R@5** |
| LoCoMo | **98.1% R@10** |
| MemBench | **98.0% R@5** |
| ConvoMem | **97.7% R@10** |
| observe | Store observation with auto-summarization, importance scoring, entity extraction, topic detection |
| recall | Retrieve verbatim content by filter (importance, type, flag, time) |
| search | Hybrid search (BM25 + vector + optional LLM judge) |
| ask | Natural language Q&A over the full memory store |
| timeline | Reverse-chronological observations with importance badges and flags |
| stats | Token economics for current session (raw vs. compressed) |
| save_knowledge | Save a knowledge entry with contradiction detection + temporal validity windows |
| search_knowledge | Search (superseded entries filtered by default) |
| promote_knowledge | Promote to global cross-project store |
{
"mcpServers": {
"context-mem": {
"command": "npx",
"args": ["-y", "context-mem"]
}
}
}Add to claude_desktop_config.json, then restart Claude Desktop.
A knowledge graph your assistant keeps between sessions — entities, relations and observations that persist.
Kill hallucinated APIs — version-accurate, up-to-date library documentation injected straight into context.
Your workspace, on speaking terms with AI — search, read and write Notion pages and databases.
A structured scratchpad for hard problems — stepwise reasoning with revisions, branches and visible logic.
Symbol-level code navigation, refactoring and memory for coding agents — the IDE brain your assistant has been missing.
Chat with your second brain — search, read and write vault notes through the Local REST API.