Log inSubmit server
Bantamkit MCP logo

Bantamkit MCP

Offline MCP memory with persistent project and profile stores, schema validation, transcript accounting, audits, and shift-work tools.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Bantamkit MCP

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "bantamkit-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "bantamkit-mcp@latest"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Automated check passed— started and listed 16 tools correctly (2d ago).

Overview

bantamkit-mcp MCP server gives coding agents persistent memory backed by local Markdown files, with project and profile stores that survive across sessions. Its tools save, recall, compact, and maintain memories, while also covering JSON Schema validation, transcript-based token accounting, skill-catalogue audits, and multi-step shift handoffs. Install it through npm or PyPI and connect it to a stdio-compatible host such as Claude Code, Claude Desktop, Cursor, or VS Code Copilot. Reach for it when an agent needs durable context and local operation after the initial installation.

Use cases

  • Save project decisions for later coding sessions
  • Recall personal preferences across agent conversations
  • Validate agent output against JSON Schemas
  • Audit skills and transcript token usage
  • Plan and hand off multi-step shift work

Key features

  • Persistent Markdown memory
  • Project and profile memory layers
  • Memory deduplication and compaction
  • JSON Schema validation
  • Transcript-based token ledger
  • Shift planning and handoffs

Tools16

What Bantamkit MCP lets your agent do. Expand a tool for its parameters.

Verified live~11.7k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

6/6 checks healthy over the last 20d

What it can do

From the 16 tools this server returned when we ran it; 16 declare their own safety hints.

Runs commands · 4Can delete or modify data · 4Handles credentials · 3Makes network requests · 1Read-only tools · 9
Network at startup3 connections
Files written at startupNone
Install scriptsNone
npm provenanceNot published
Package age60 days
Maintainers1

Observed 2026-11-04 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 16d ago via OSV.dev · bantamkit-mcp@latest (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars2Stargazers on the source repository.
npm downloads1.6kPackage downloads in the last 30 days.
Last commitTodayMost recent push to the default branch.
Availability100%Our rolling endpoint + install checks that succeeded.
Install checkPassedOur sandbox started it and listed its tools.
Tools exposed16Callable tools this server registers over MCP.

How Bantamkit MCP works

What bantamkit-mcp does

bantamkit-mcp adds durable local memory to coding agents. Agents can preserve preferences, project decisions, and solutions, then retrieve them in later sessions instead of rebuilding the same context. Memory is stored as Markdown beneath .bantamkit/memory, with separate project and profile locations available.

The server also covers tasks beyond recall. Its capabilities include JSON Schema validation, a token ledger derived from host transcripts, skill-catalogue auditing, and shift-work operations for planning and handing off multi-step jobs. A matching Python library is available for small-model agents, while the MCP implementations for Node and Python share the same on-disk store.

How it works

Memory writes target the project store. Recall can read the project store, stores explicitly granted through .bantamkit/config.yaml, and the profile store at ~/.bantamkit/memory. Saved entries are deduplicated and kept within a byte budget. Compaction and dreaming tools help maintain the memory collection over time.

The server communicates with hosts over MCP using stdio. The npm and PyPI routes both support an install-once setup that records an absolute executable path, allowing later launches to run without network access or dependence on the shell’s PATH. Claude Code can optionally use hooks to add the memory index at session start, surface relevant facts on prompts, and encourage a save after a long session with no saved memory.

Setup and configuration

For a Node installation, run npx -y bantamkit-mcp@latest --install <host>, replacing <host> with claude, claude-desktop, copilot, or cursor. The installer writes the host entry, backs up files it changes, and avoids prompting. Python users can install the MCP extra into a persistent virtual environment with pip install "bantamkit[mcp]", then run bantamkit-mcp --install <host>.

The host must be restarted after installation; Claude Code requires a new session. An alternative configuration launches npx -y bantamkit-mcp at every session, but that route checks the package registry on each launch and can hang when offline. Install-once mode is better suited to machines that must operate without network access.

Tools and capabilities

  • memory_save stores facts in the project memory.
  • memory_recall retrieves relevant saved information.
  • memory_compact and memory_dream maintain the memory store.
  • JSON Schema validation checks structured data.
  • A token ledger reads usage information from host transcripts.
  • A skill-catalogue audit reviews available skills.
  • Shift-work tools plan and hand off multi-step jobs.

Limitations and notes

The Node and Python implementations share files, so both runtimes should point at the intended stores. Install-once entries record the executable path used during setup; changing Node versions requires reinstalling or forcing the installation. The server is designed for stdio hosts, and its offline behavior begins after the package has been installed locally.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Bantamkit MCP

bantamkit-mcp is an MCP server that connects coding agents such as Claude Code, Claude Desktop, Cursor, and VS Code Copilot to persistent local memory. Its main tools save, recall, compact, and maintain memories, with additional support for JSON Schema validation, transcript token accounting, skill audits, and shift-work handoffs.

Install it with `npx -y bantamkit-mcp@latest --install <host>` for Node or `pip install "bantamkit[mcp]"` followed by `bantamkit-mcp --install <host>` for Python. Supported installer host names include `claude`, `claude-desktop`, `copilot`, and `cursor`.

Yes, bantamkit-mcp supports Cursor through the `--install cursor` setup route. The installer writes the MCP entry and backs up the file it edits; restart Cursor after installation so it can launch the server.

Yes, after an install-once setup, bantamkit-mcp launches from an absolute local path without the network. The repeated `npx` configuration checks the registry at each launch, so it is not suitable for reliably offline startup.

Related MCP servers

All in Knowledge & Memory All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewBantamkit MCP AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/bantamkit-mcp?style=directory)](https://allmcps.com/mcp/bantamkit-mcp)
HTML Embed
<a href="https://allmcps.com/mcp/bantamkit-mcp"><img src="https://allmcps.com/api/badge/bantamkit-mcp?style=directory" alt="Bantamkit MCP on AllMCPs" /></a>