MCP.Market

Context Mem MCP server

by JubaKitiashvili·io.github.JubaKitiashvili/context-mem·v3.2.1·19 stars
No reviews yet

Persistent memory for AI agents — 98%+ retrieval recall, 99% token savings, 44 tools

C63/100grade C
Write a reviewInstall
What users say
No reviews yet
Be the first
Safety scan
C63/100

full report

Adoption
Growing

19 stars179 downloads/wk

Reviews

Write one

Nobody has reviewed Context Mem yet.

If you have run it, two minutes of your experience saves the next person an afternoon.

Context Mem tools (44, 3 write)

write = sends, deletes, buys or posts

Read from the package source without running it. The installed server may list more.

  • add_relationship

    Add a relationship between two entities in the knowledge graph. Creates entities if they do not exist.

  • agent_broadcast

    Broadcast a message to all active agents via the event system.

  • agent_register

    Register the current session as a named agent for multi-agent coordination.

  • agent_status

    List all active agents with their tasks and claimed files.

  • ask

    Ask a natural language question about the project. Searches knowledge, observations, events, and graph entities.

  • browse

    Browse observations by topic, person, or time dimension.

  • budget_configure

    Configure budget settings (session limit, overflow strategy).

  • budget_status

    Get current budget usage and status for the session.

  • claim_files

    Claim files for the current agent. Returns conflicts if already claimed by another agent.

  • configurewrite action

    Update a mutable configuration key at runtime.

  • emit_event

    Emit a context event with priority and metadata.

  • entity_detect

    Extract entities (technologies, people, files, components) from text content.

  • executewrite action

    Execute code in JavaScript, TypeScript, Python, Shell, Ruby, Go, Rust, PHP, Perl, R, or Elixir.

  • explain_decision

    Reconstruct the evidence chain behind a code change or decision. Returns the trail of events that led to a decision.

  • find_tunnels

    Find topics that appear in 2+ projects (cross-project knowledge bridges).

  • generate_story

    Generate a human-readable narrative from session data. Formats: pr (pull request), standup, adr (architecture decision record), onboarding.

  • get

    Retrieve a single observation by its ID.

  • global_search

    Search the global cross-project knowledge store.

  • graph_neighbors

    Find all entities connected to a given entity. Faster than graph_query for simple lookups.

  • graph_query

    Query the knowledge graph. Find entities and their relationships.

  • handoff_session

    Generate session handoff — saves state and returns continuation prompt for a new session. Use when context is running low or before ending a session.

  • import_conversations

    Import external conversation exports (Claude, ChatGPT, Slack, plaintext) into context memory.

  • index_content

    Index content into the content store with code-aware chunking for later search.

  • list_people

    List all detected person entities with relationship counts.

  • list_topics

    List all detected topics with observation counts.

  • merge_suggestions

    View pending merge suggestions for duplicate global knowledge entries.

  • observe

    Store a new observation (content snippet) into context memory with automatic summarization.

  • predict_loss

    Predict which memory entries are at highest risk of being forgotten or archived. Users can pin important entries to protect them.

  • promote_knowledge

    Promote a project knowledge entry to the global cross-project knowledge store. Privacy engine sanitizes content before storing.

  • query_events

    Query context events with optional filters.

  • recall

    Verbatim memory retrieval with importance filtering and rich attribution. Returns original content, not summaries.

  • resolve_contradiction

    Resolve a contradiction between knowledge entries by merging, superseding, or keeping both.

  • restore_session

    Restore a previous session snapshot by session ID.

  • save_knowledge

    Save a knowledge entry with automatic contradiction detection. When contradictions are found, the save is blocked — resubmit with force: true to save anyway.

  • search

    Search stored observations using BM25 + trigram fusion.

  • search_content

    Search the content store for indexed content chunks.

  • search_knowledge

    Search the knowledge base using 3-layer search (FTS5 → trigram → scan). Optionally include global cross-project knowledge. By default only returns currently-valid facts (valid_to IS NULL).

  • stats

    Get token economy statistics for the current session.

  • summarize

    Summarize a piece of content using the matching summarizer plugin without storing it.

  • temporal_query

    Query knowledge that was valid at a specific point in time. Returns facts that were active at the given timestamp.

  • time_travel

    View or compare the project state at any point in time. Shows observations, knowledge, and events as of a target date.

  • timeline

    Retrieve observations in reverse-chronological order with optional filters.

  • update_profilewrite action

    Update the project quick profile — a 3-5 line summary shown at every session start. Auto-generates from knowledge if no content provided.

  • wake_up

    Generate a scored session primer with 4-layer context: project profile, critical knowledge, recent decisions, and top entities.

Public scan report

scanner v0.1.9 · 2026-09-27 · same rubric, same numbers if you re-run it

1 high
  • Code scan247 source files scanned13/25
  • –Live reliabilityno gateway calls yet and no remote to proben/a
  • –Tool poisoningtools not inspected (local package is not executed); not countedn/a
  • Auth qualitylocal package, no credentials required12/15
  • Maintenancelast push 161 days ago8/15
  • Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Findings (1)

  • highShell command built from a string (injection risk)exec.shell-concat
    hooks/session-start-hook.js: …date-check.js'); const updateResult = execSync(`node "${updateScript}" 2>/dev/null`, { timeo…
Overall 63/100. Components that don't apply are left out of the denominator. Any critical finding is an F.RubricAppeal a findingJSON

How Context Mem compares

Against every graded MCP server, from our own scans and uptime checks. Updated after each scan run.

Safety score
63/100
higher than 18% of 48,255 servers
Tools
44
the typical server has 7
Tools that change things
7%
typical: 0%. Fewer write tools means less an agent can break
GitHub stars
19
more than 91% of 28,688 servers
Downloads a week
179
more than 73% of 12,191 servers with a package

What its tools let an agent do: 36 read · 4 create · 3 change · 1 run or deploy

Install Context Mem in Claude Code, Cursor or VS Code

Runs npx -y context-mem on your machine. Read the scan report first; the gateway never runs local packages.

claude mcp add context-mem -- npx -y context-mem
Add to Cursor

What the publisher says

From the Context Mem repository's README, as published. We do not edit it. Read it on GitHub

<!-- Banner placeholder: dual-pillar visual — memory wiki page + compression savings chart -->

Context Mem

Memory + context infrastructure for AI agents. Remembers everything. Compresses everything. Fully local.

-gold)

The Problem

Two problems with today's AI tooling that no one has solved together in a single package.

Your AI forgets. Every new session starts from zero. The architecture decisions you settled on last Thursday, the bug you spent four hours tracing to a misconfigured environment variable, the preferences you stated three times — none of it carries forward. You spend the first ten minutes of every session re-explaining context that already existed. Multiply this by every developer on your team, every project, every day.

Your context explodes. Long coding sessions blow past the context window. A typical session with 50 tool outputs accumulates 365 KB of raw text — stack traces, test output, file reads, shell commands. Every token costs money or slows the model. Naive truncation drops the exact evidence the model needs. Keeping everything makes responses slower and inference cost climb fast.

These two problems compound each other. The solution to forgetting (keep everything) is the opposite of the solution to context explosion (discard everything). The result is a false tradeoff most tools force on you: either your AI forgets everything, or your costs balloon. context-mem solves both simultaneously by building an indexed, compressed, retrievable memory store rather than dumping raw history into the context window.

The Solution — one tool, two pillars

Pillar 1: Memory (LLM Wiki)

Every tool call is automatically ingested, summarized, and written into a navigable markdown vault — a living wiki your AI maintains about your project. Entities get their own pages with backlinks. Topics get synthesis pages. Sessions become browseable source documents. Decisions accumulate into a reconstructible trail.

The vault lives at .context-mem/vault/ and syncs continuously from the underlying SQLite store. Read it in Obsidian, grep it from the terminal, or query it through 45+ MCP tools using hybrid BM25 + vector + optional LLM judge search. The raw SQLite store is the authoritative record; the markdown vault is the derived, human-readable layer.

This is a reference implementation of Andrej Karpathy's LLM Wiki pattern — three layers (raw sources / wiki / schema), with automatic ingest from tool calls that no other system provides.

Pillar 2: Compression (14 summarizers)

Every observation passes through a content-aware summarizer before storage. A stack trace is not treated the same way as a JSON config file. Shell output from a build is compressed differently from TypeScript compiler errors. The system applies the right compression for the content type.

The result: a full coding session with 50 tool outputs goes from 365 KB to 3.2 KB — 99.1% token savings, verified. Compression is adaptive: recent high-importance observations stay verbatim; older low-importance ones compress progressively. Pinned entries never compress regardless of age.

One Command

npm i context-mem && npx context-mem init

init auto-detects your editor and writes the right config files:

Shortened. The full README is on GitHub.

Nothing above is checked by us. What we check is on the safety report.

Context Mem: common questions

Is Context Mem MCP server safe?
With care: it is graded C, so read the findings first (63/100). Read the Context Mem safety report
How do I install Context Mem?
It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.
Does Context Mem need an API key?
Not as far as the registry entry and our scan can tell: no credentials are declared or required.
Is Context Mem maintained?
The last commit was 199 days ago (2026-04-19). The latest release is v3.2.1.
What can I use instead of Context Mem?
Servers from other publishers that do the same job: Codebase Memory MCP server, Honcho MCP server and State Memory MCP server.
How does Context Mem compare with other MCP servers?
safety score 63/100 (higher than 18% of 48,255 servers); tools 44 (the typical server has 7); tools that change things 7% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.

Alternatives to Context Mem

Same job from other publishers: the closest match first, then the best rated.

  • Codebase Memory
    Codebase knowledge graph for AI agents — 162 languages, sub-ms queries, 99% fewer tokens.
    C
  • Honcho
    Memory that reasons: continual learning for stateful agents. Better context, fewer tokens.
    A
  • State Memory MCP
    Persistent workflow graph server tracking tasks, decisions, spatial entities, and blockers.
    A
  • Knowl
    Persistent memory for AI agents, for code, research and everyday life. Stale facts retire.
    A