{"slug": "llm-kosh", "title": "LLM KOSH", "summary": "LLM KOSH, a local-first memory cartridge for MCP-compatible AI clients, has been released. The tool provides durable, inspectable memory storage without relying on hosted services, featuring a read-only MCP server, CLI, and background service. It is available now for Python 3.10+ with packaging for Windows, macOS, and Linux in progress.", "body_md": "`llm-kosh`\n\nis a local-first memory cartridge for MCP-compatible AI clients.\nIt gives your agents durable memory without handing your workspace to a hosted\nmemory service.\n\nThink of it as a structured, inspectable memory layer for agents:\n\n- plain files you can back up, diff, and review\n- a tamper-evident ledger for every mutation\n- a read-only-by-default MCP server\n- a background service for intake and maintenance\n- a CLI for local control and automation\n\n- Keep AI context local and auditable.\n- Separate the cartridge root from the repository root.\n- Drop receipts or intake files into watched folders and let the service absorb them.\n- Connect MCP clients with minimal privilege by default.\n- Publish and verify the same artifact through GitHub Actions.\n\nThe core project is usable now:\n\n- the CLI runs locally\n- the Python package installs and works\n- the MCP server runs locally\n- the service can watch intake folders\n- the GitHub Actions publish path is working\n\nThe remaining work is release polish for Windows, macOS, and Linux packaging.\n\nPython 3.10 or newer is required.\n\n```\npython -m pip install --upgrade llm-kosh\nllm-kosh install --yes\nllm-kosh status\n```\n\nThat installs the package, creates the default cartridge at\n`~/.llmkosh/cartridge`\n\n, configures local defaults, and registers the supported\ndesktop integration where possible.\n\nTo manage the background service:\n\n```\nllm-kosh service start\nllm-kosh service status\nllm-kosh service stop\n```\n\nIf you want to work in a custom cartridge location, set the root explicitly:\n\n```\nllm-kosh --root ./my-cartridge init --owner \"Local User\"\nllm-kosh --root ./my-cartridge add --kind note --title \"First memory\" --body \"Hello\"\nllm-kosh --root ./my-cartridge query \"Hello\"\n```\n\nThere are three folders worth knowing:\n\n- the repository root: the code checkout you are reading now\n- the cartridge root: the live memory store selected by\n`--root`\n\nor`LLMKOSH_ROOT`\n\n- watched intake folders:\n`receipts/`\n\n,`intake/`\n\n, and any configured external drop folders\n\nIf you drop files into the cartridge’s intake areas, the service can process\nthem asynchronously. If you configure external folders through\n`[daemon].watched_directories`\n\n, the service can absorb those too.\n\n```\nllm-kosh --root ./my-cartridge mcp-server\n```\n\nThe MCP server starts read-only.\n\nEnable stronger capabilities only for clients that should be allowed to write, mutate, or export private context:\n\n```\nllm-kosh --root ./my-cartridge mcp-server --allow-write\nllm-kosh --root ./my-cartridge mcp-server --allow-write --allow-mutate\nllm-kosh --root ./my-cartridge mcp-server --allow-private\n```\n\nYou can also run MCP over local HTTP:\n\n```\nllm-kosh --root ./my-cartridge mcp-server --http --port 8000\n# endpoint: http://127.0.0.1:8000/mcp\n```\n\n- Python CLI for creating, searching, packing, importing, and verifying cartridges\n- read-only-by-default MCP server\n- local background service for intake and maintenance jobs\n- optional desktop packaging with a bundled CLI sidecar\n- plain-file storage that stays inspectable, backupable, and Git-friendly\n- optional extras for filesystem watching, service integration, semantic search, and ingest helpers\n\n```\npython -m pip install \"llm-kosh[watch]\"     # filesystem events\npython -m pip install \"llm-kosh[server]\"    # FastAPI service\npython -m pip install \"llm-kosh[semantic]\"  # local vector search\npython -m pip install \"llm-kosh[ingest]\"    # document conversion helpers\npython -m pip install \"llm-kosh[all]\"       # all optional features\n```\n\nMCP support is included in the base installation.\n\n```\npython -m pip install -e \".[server,watch,ingest]\"\npython -m pytest -q\n```\n\nIf you are changing packaging or release behavior, also run:\n\n```\npython -m build\npython -m twine check dist/*\n```\n\n- Storage and search are local by default.\n- There is no automatic cloud sync or telemetry in the Python package.\n- MCP starts read-only.\n- Write, mutation, and private-export capabilities require explicit opt-in.\n- Context exports are checked for common secret patterns before sharing.\n- Cartridge files are plaintext; use operating-system disk encryption if local data at rest needs encryption.\n\nSee [SECURITY.md](/rastogivaibhav/llm-kosh/blob/master/SECURITY.md) and [docs/SECURITY.md](/rastogivaibhav/llm-kosh/blob/master/docs/SECURITY.md) for\nboundaries and limitations.\n\nThe Electron desktop app is packaged separately from the Python package. Local developer builds and Windows installer smoke tests are supported. Public GA desktop distribution still requires verified Windows code signing and macOS Developer ID signing/notarization.\n\nFor the current release posture across package, MCP, service, and desktop,\nsee [GA_READINESS.md](/rastogivaibhav/llm-kosh/blob/master/GA_READINESS.md).\n\n[Quickstart](/rastogivaibhav/llm-kosh/blob/master/QUICKSTART.md)[Architecture](/rastogivaibhav/llm-kosh/blob/master/docs/ARCHITECTURE.md)[CLI reference](/rastogivaibhav/llm-kosh/blob/master/docs/CLI_REFERENCE.md)[MCP guide](/rastogivaibhav/llm-kosh/blob/master/docs/MCP_GUIDE.md)[Developer guide](/rastogivaibhav/llm-kosh/blob/master/docs/DEVELOPER_GUIDE.md)[Developer FAQ](/rastogivaibhav/llm-kosh/blob/master/docs/DEVELOPER_FAQ.md)[MCP developer guide](/rastogivaibhav/llm-kosh/blob/master/docs/MCP_DEVELOPER_GUIDE.md)[Service developer guide](/rastogivaibhav/llm-kosh/blob/master/docs/SERVICE_DEVELOPER_GUIDE.md)[Desktop developer guide](/rastogivaibhav/llm-kosh/blob/master/docs/DESKTOP_DEVELOPER_GUIDE.md)[Release engineering](/rastogivaibhav/llm-kosh/blob/master/docs/RELEASE_ENGINEERING.md)[Documentation standards](/rastogivaibhav/llm-kosh/blob/master/docs/DOCUMENTATION_STANDARDS.md)[GA readiness](/rastogivaibhav/llm-kosh/blob/master/GA_READINESS.md)[Archived historical docs](/rastogivaibhav/llm-kosh/blob/master/docs/archive/README.md)\n\nNative C++ math acceleration is optional. Set `LLM_KOSH_BUILD_NATIVE=1`\n\nand\ninstall `pybind11`\n\nbefore building if you want to test it. Release wheels use\nthe portable pure-Python fallback.\n\nLicensed under the MIT License.", "url": "https://wpnews.pro/news/llm-kosh", "canonical_source": "https://github.com/rastogivaibhav/llm-kosh", "published_at": "2026-06-30 15:38:03+00:00", "updated_at": "2026-06-30 15:49:44.677866+00:00", "lang": "en", "topics": ["ai-tools", "developer-tools", "ai-agents", "ai-infrastructure", "large-language-models"], "entities": ["LLM KOSH", "MCP", "GitHub Actions", "Python", "FastAPI"], "alternates": {"html": "https://wpnews.pro/news/llm-kosh", "markdown": "https://wpnews.pro/news/llm-kosh.md", "text": "https://wpnews.pro/news/llm-kosh.txt", "jsonld": "https://wpnews.pro/news/llm-kosh.jsonld"}}