About Mnemoscope
Mnemoscope’s MCP server provides local tools for assessing context rot, reading a Markdown vault by freshness tier, and maintaining a signed journal. It is for MCP clients that work with vaults and keeps vault reads and writes on the local machine.
Key features
- 01Score context-rot risk
- 02Split vault reads by freshness tier
- 03Record signed journal entries
- 04Read journal entries by session
- 05Keep vault data local
Use cases
- 01Assess context-rot risk in a vault
- 02Read working, episodic, and semantic notes by freshness
- 03Record signed file operations
- 04Review journal history for a session
MCP support
The server exposes context-rot scoring, tiered reading, and signed-journal tools to MCP-compatible clients over stdio. Reads and writes remain local; it does not transmit vault content.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
4Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| vault_path | string | Yes | Absolute path to the vault root directory. |
Details
From the project README@mnemoscope/mcp-server
MCP server exposing Mnemoscope's context-rot scoring, tiered reading, and signed journal tools to any Model Context Protocol–compatible client (Claude Code, Cursor, ChatGPT desktop, anything that speaks MCP over stdio).
100% local. No cloud. No telemetry. Apache-2.0.
Install
npm install -g @mnemoscope/mcp-server
# the `mnemoscope-mcp` binary is now on your PATH
Or run via npx without a global install:
npx -y @mnemoscope/mcp-server
Connect to Claude Code
// ~/.claude/settings.json
{
"mcpServers": {
"mnemoscope": {
"command": "npx",
"args": ["-y", "@mnemoscope/mcp-server"]
}
}
}
Restart Claude Code. The four tools below are discovered automatically.
Tools
| Tool | Input | What it returns |
|---|---|---|
predict_rot | vault_path | 0–100 risk score, dominant factor, full factor breakdown, top-5 risk notes, vault stats |
get_tiered_read | vault_path, optional working_max_age_days, episodic_max_age_days | Working / episodic / semantic split of the vault by freshness |
record_journal | vault_path, session_id, op (read/write/create/delete), target_path, optional content | The signed entry, including its sig, keyFingerprint, and prevHash |
read_journal | vault_path, optional session_id | All journal entries, or a single session's |
Example — predict_rot
{
"rot_risk": 41,
"dominant_factor": "tokenVolume",
"factors": {
"tokenVolume": 100, "semanticRedundancy": 0,
"distractorDensity": 2.65, "structuralCoherence": 100, "freshnessSpread": 0
},
"top_risk_notes": [
{ "relPath": "brainstorms/.../transcript.md", "approxTokens": 13439, "reason": "very large note" }
],
"vault_stats": { "noteCount": 113, "approxTokens": 506823 },
"baseline_model": "v0-heuristic",
"version": "0.1.0"
}
Bootstrap a vault
Before the journal tools work the vault needs an Ed25519 keypair. The companion @mnemoscope/cli ships mnemoscope-init for that:
npx -p @mnemoscope/cli mnemoscope-init /path/to/your/vault
The auto-onboarding modal in the Obsidian plugin does the same on first launch.
Verify the journal
npx -p @mnemoscope/cli mnemoscope-verify /path/to/your/vault
Exits non-zero on field tampering, hash-chain break, or entries signed by a key the current vault does not own.
What this server does NOT do
- It does not store or transmit your vault content. All reads are local; all writes are local.
- It does not call any external service. (The optional
mnemoscope-timestampCLI in@mnemoscope/cliPOSTs only a SHA-256 digest of each journal entry's signature to a public OpenTimestamps calendar — never any vault content.) - It does not replace a memory store. It is an instrument for one — see the README for the comparison with Letta, Mem0, Zep, Cognee, MemPalace, and Smart Connections.
Repo + docs
- Source, issues, releases: https://github.com/toonight/Mnemoscope
- Architecture, threat model, scientific posture: see the main README
License
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.