Skip to content
Agentmart
Submit a tool

Memory Kernel

Local-first memory for AI agents: SQLite FTS5, deterministic recall, no vector DB, no cloud.

Connector by Artem362 · GitHub account verified by the MCP Registry

IndexedApache-2.0Local (stdio)★ 3
Write a review

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/artem362-memory-kernel

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Local-first memory for AI agents: SQLite FTS5, deterministic recall, no vector DB, no cloud.

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools
5 read · 3 write · 0 delete
Secrets
None asked for
Install scripts
Not scanned yet
Runs as
Python package

Tools by what they do 8

Read from the source code; the running server may list different tools.

Read 5 · Look things up without changing anything.

  • memory_build_context
  • memory_list
  • memory_search
  • memory_stats
  • memory_wake_up

Write 3 · Create, change, send or run things.

  • memory_forget
  • memory_ingest
  • memory_remember

Secrets and settings it asks for

Other settings (environment variables or headers).

  • MEMORY_KERNEL_DB

Files it touches

  • Any path the agent passes to its tools

Tools (8)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • memory_build_contextBuild a context packread-only

    Assemble a compact, budget-limited context pack for the current task. Returns a single rendered block sized to fit ``budget_chars`` that you can drop straight into a prompt. Near-duplicate memories are dropped so the pack carries distinct signal. Prefer this over raw search when you want a ready-to-inject summary rather than a list to reason over. Returns a JSON object with 'budget_chars', 'used_chars', 'items', and 'rendered' (the text block to inject). On failure: "Error: <message>".

  • memory_forgetArchive (soft-forget) a memory

    Archive a memory so it stops surfacing in recall, while keeping it recoverable. This is a soft forget, not a delete: the memory is hidden from search, context, wake-up, and list, but the data is preserved and a human can bring it back from the CLI (`memory-kernel restore`). Use this when a memory is stale or no longer relevant. Hard deletion is not available over MCP. Returns a confirmation, or a not-found message. On failure: "Error: <message>".

  • memory_ingestIngest raw text into structured memories

    Split raw notes or a transcript into structured memories and save them. Use this to capture a chunk of conversation, meeting notes, or planning text without structuring it yourself. Each segment gets an inferred kind, title, tags, importance, and certainty. Deduplication applies per segment. Returns a summary of how many memories were stored (created vs updated) followed by one line per memory. On failure: "Error: <message>".

  • memory_listList recent memoriesread-only

    List recent memories (most recently updated first) with optional filters. Use this to browse what is stored rather than to answer a specific query. Archived and superseded memories are excluded. Returns a markdown list or JSON array of memory records. On failure: "Error: <message>".

  • memory_rememberSave a precise memory

    Save one well-formed memory to the local store. Use this when you already know exactly what to remember (a decision, a constraint, a user preference, a fact). The store deduplicates: saving the same memory again merges into the existing record, so repeated calls are safe. Returns a confirmation with the stored memory id and whether it was 'created' or 'updated'. On failure: "Error: <message>".

  • memory_searchSearch memoriesread-only

    Find the few memories most relevant to a query. Ranking is deterministic (lexical match, importance, certainty, recency, access frequency). Ukrainian inflections are bridged: a query like 'рішення' also matches stored 'вирішили'. Archived and superseded memories are excluded. Returns a markdown list (title, kind/scope, id, excerpt, metadata) or a JSON array of full memory records. On failure: "Error: <message>".

  • memory_statsMemory store statisticsread-only

    Report store statistics: total count, breakdown by kind and scope. Pass ``since`` ('7d' or an ISO date) to also get recent-activity counts. Returns a markdown summary or JSON object with totals and breakdowns. On failure: "Error: <message>".

  • memory_wake_upBuild a wake-up packread-only

    Build a small 'hot memory' pack to load at the start of a session. Returns the most important/certain/recent memories with no query, sized to ``budget_chars``. Use this once when a session begins to prime the agent. Returns a JSON object with 'budget_chars', 'used_chars', 'items', and 'rendered'. On failure: "Error: <message>".

Signals

Stars
3
Forks
0
Open issues
0
Weekly downloads
45
Last commit
Jun 15, 2026 (4 months ago)
Last release
Jun 12, 2026 (4 months ago)
Last package publish
Jun 15, 2026 (4 months ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name
io.github.Artem362/memory-kernel
Version
0.3.2
Auth
No auth
Packages
pypi amormorri-memory-kernel
License
Apache-2.0
Repository
github.com/Artem362/memory-kernel
First seen
Oct 24, 2026
Last checked
Oct 25, 2026
Source
Official MCP Registrygithub-graphql
Cite this page

Memory Kernel MCP server. Agent Mart. https://agentmart.dev/connectors/artem362-memory-kernel (last updated 2026-10-25)