Toon Memory MCP server

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

• context_brief

  Generate a compact context briefing: relevant memory + active sessions + project health. Single call instead of 5-6 separate calls. Zero LLM, pure deterministic logic.

• context_diff

  What changed since the last session: git commits + modified files + new/updated memory entries. Zero LLM.

• context_export

  Export all memory as markdown injectable for other agents or sessions. Full (detailed) or compact format. Zero LLM.

• context_focus

  Hyper-focused context for a specific task: relevant entries + related files + code referencing the symbol + existing tests. Zero LLM.

• context_generate

  Generate a full system prompt: project (package.json, deps, framework) + git (branch, commits) + memory (relevant entries) + sessions. Single call to prepare the agent. Zero LLM.

• context_health

  Full memory health audit: orphan links, duplicates, broken refs, expired TTL, quality, stale sessions. Includes score 0-100. Zero LLM.

• memory_archive

  Move old entries (>30 days) to archive.toon to keep memory clean.

• memory_backupwrite action

  Create a timestamped backup of the memory file. Use before major operations or to preserve state.

• memory_captured

  Show activity log captured automatically by hooks (only if capture is enabled). Useful for promoting observations to memory with memory_remember.

• memory_checkpoint

  Save or update a session checkpoint with 7-day TTL. Creates a knowledge entry snapshot of the current session state. Same session = upsert (single checkpoint per session). Auto-expires.

• memory_compress

  LLM-powered compression: select entries to compress into one concise entry. The agent (YOU) reads the entries, writes a compressed summary, then calls memory_remember to save it. Two-step dance: this tool provides context, you provide intelligence.

• memory_consolidate

  Cleanup operations, all deterministic (no LLM): identical removes duplicate entries (keeps the first), similar merges entries with >50% word overlap (keeps longer content, combines tags), low-quality removes entries below minQuality or without tags, versions marks entries describing the same subject with older versions obsolete in favor of the newest.

• memory_decrypt

  Disable encryption and decode memory.

• memory_diff

  Show what changed in memory since a date. Useful for seeing what was learned since the last session.

• memory_encrypt

  Enable AES-256-GCM encryption for memory. Requires TOON_MEMORY_KEY in environment.

• memory_export_gist

  Export memory to a private GitHub Gist for cloud sync. Requires GITHUB_TOKEN env var. Zero deps \u2014 uses Node's built-in fetch().

• memory_export_globalwrite action

  Write the current project memory to the global memory file (~/.toon-memory/memory/global.toon, or TOON_MEMORY_GLOBAL_FILE). One-shot: the global file is a plain TOON document you curate and import into other projects with memory_import_global.

• memory_forget

  Lifecycle operations for a memory entry: soft (default) marks it obsolete and hides it from recalls, hard permanently removes it, restore brings it back to active, supersede retires it with a typed superseded_by edge to a newer entry.

• memory_graph_path

  Find the shortest path between two memory entries in the knowledge graph. Shows how two concepts are connected through related entries.

• memory_import_gist

  Import memory from a GitHub Gist. Merges entries (keeps newer dates, combines tags). Requires GITHUB_TOKEN env var.

• memory_import_global

  Merge cross-project conventions from the global memory file (~/.toon-memory/memory/global.toon, or TOON_MEMORY_GLOBAL_FILE) into this project. One-shot merge by key: keeps newer dates, unions tags/links, preserves all fields. Deterministic and offline.

• memory_merge_sessions

  Merge observations from multiple sessions/branches into a consolidated view. Deduplicates and suggests entries to promote to memory. Inspired by Hindsight's multi-agent session merging.

• memory_pin

  Pin a memory entry with priority level (1-5). Higher priority entries appear first in recalls. Default: 1. Useful for critical decisions, project rules, or frequently-needed context.

• memory_primer

  Generate a compact system primer: always-current knowledge map for auto-injection into agent context. Excludes private entries, shows top memories by importance, categories, and patterns. Use at session start or when context feels stale.

• memory_promote

  Heuristic auto-promote from the capture log (no LLM). Each observation gets a deterministic confidence score; high-confidence ones promote as active entries, low-confidence (\u2264 minConfidence) become reviewable drafts. Opt-in: run with dryRun: true first.

• memory_recall

  Search the project's persistent memory. Returns relevant entries. Use BEFORE reading files. Pattern: recall \u2192 context_diff for recent changes \u2192 capture to save observations.

• memory_reflect

  Deterministic memory audit (no LLM): tag co-occurrence themes, hub/orphan entries, near-duplicates, contradiction candidates, stale entries, and superseded-but-active entries. Signals what to link, merge, or retire.

• memory_remember

  Save a fact to the project's persistent memory (decisions, patterns, bugs, knowledge). Persists between sessions. Tags: 'private' excludes from context injection; 'superseded' marks as obsolete.

• memory_search

  Search memory with combined filters: text query, category, tags, and date range. Returns ranked results.

• memory_secret

  Store/retrieve secrets in the encrypted sidecar vault (secrets.toon, AES-256-GCM). Keeps data.toon readable and open while sensitive values stay encrypted at rest. Requires TOON_MEMORY_KEY. 'store' encrypts, 'get' decrypts, 'list' shows metadata only (no values), 'forget' removes.

• memory_sessions

  Show active agent sessions in this project (git branch, files touched, last-seen) and detect soft conflicts (files touched by 2+ sessions). Use at startup to avoid overwriting other parallel sessions' work.

• memory_smart_recall

  Unified retrieval: combines BM25 + graph + decay + quality in a single call. Use at the START of each task to get all relevant memory context. Then use context_diff for recent changes, context_focus for deep dives.

• memory_stats

  Show project memory statistics.

• memory_suggest

  Suggest memory entries related to a given context. Useful for gathering context before a task.

• memory_summary

  Save or retrieve a summary of a large file to save tokens.

• memory_tag

  Batch add, remove, or set tags on one or more memory entries by key or id.

• memory_unpinwrite action

  Remove the pin from a memory entry so it returns to normal ranking.

• Code scan60 source files scanned20/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitystatic API keys via environment variables6/15
• Maintenancelast push 50 days ago12/15
• Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Safety score

71/100

higher than 38% of 4,460 AI and agents servers

Tools

37

the typical AI and agents server has 6

Tools that change things

8%

typical: 0%. Fewer write tools means less an agent can break

GitHub stars

11

more than 85% of 2,711 AI and agents servers

Downloads a week

119

more than 67% of 1,538 AI and agents servers with a package

Toon Memory: common questions

Is Toon Memory MCP server safe?

Mostly: it is graded B (71/100). Read the Toon Memory safety report

How do I install Toon Memory?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does Toon Memory need an API key?

Yes. The registry entry asks for TOON_MEMORY_KEY, GITHUB_TOKEN.

Is Toon Memory maintained?

The last commit was 91 days ago (2026-08-05). The latest release is v4.3.5.

What can I use instead of Toon Memory?

Servers from other publishers that do the same job: Total Agent Memory MCP server, Memanto MCP server and Serena MCP: the IDE for your agent MCP server.

How does Toon Memory compare with other AI and agents MCP servers?

safety score 71/100 (higher than 38% of 4,460 AI and agents servers); tools 37 (the typical AI and agents server has 6); tools that change things 8% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.