MemKraft: memory tool for Hermes Agent

Local-first plain-Markdown knowledge system for AI agents, usable as a Hermes Agent memory provider

What MemKraft does

MemKraft keeps knowledge in plain Markdown: entity pages, decisions, timelines and notes you can read, diff, edit and version, with local JSONL sidecars for operational records such as canonical events and outcomes. The core install uses the Python standard library, makes no model calls and needs no API key; your agent supplies the intelligence while MemKraft supplies storage, provenance, lifecycle controls and a feedback ledger. Canonical facts require a source, and retrieval preserves source links.

Memory is tied to what happens after recall: compile_context() returns a stable usage_id, and report_outcome() records success or failure so later context ordering adjusts deterministically. Governance covers forgetting, do-not-remember policies, tombstones, dry-run lifecycle operations, audit logs and fail-closed reads. It is used through a CLI, a Python API or an MCP server, and the README says Hermes Agent users can enable it as the active memory provider, with verified versions and profile-safe steps in docs/HERMES_AGENT.md.

Key features

Plain-Markdown entity pages, decisions, timelines and notes, with local JSONL sidecars for events and outcomes
Core install uses only the Python standard library, with no model calls and no API key
Canonical facts require a source, and retrieval keeps source links
compile_context and report_outcome loop that adjusts later context ordering from recorded results
Governance tools including do-not-remember policies, tombstones, dry-run lifecycle operations and audit logs
CLI, Python API and MCP server interfaces

When to use it

Using MemKraft as the active memory provider for a Hermes Agent profile
Keeping agent knowledge in files you can diff and version
Recording which retrieved context helped so later retrieval improves

Who it is for: Developers who want accountable, file-based agent memory that works across agents and models.

How it fits with Hermes Agent

Hermes Agent users can enable MemKraft as the active memory provider; the README points to docs/HERMES_AGENT.md for verified versions and profile-safe setup steps.

How to install MemKraft

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

pipx install memkraft
memkraft init

Requirements: Python 3.9+.

FAQ

What is MemKraft?

MemKraft is a local-first, plain-Markdown knowledge system for AI agents. It stores sourced facts, retrieves bounded context and records outcomes so later retrieval can improve.

Does MemKraft work with Hermes Agent?

Yes, the README says Hermes Agent users can enable MemKraft as the active memory provider. The exact verified versions and profile-safe setup steps are in docs/HERMES_AGENT.md.

How do I install MemKraft?

Run pipx install memkraft, then memkraft init, which creates a ./memory/ directory by default. You can also use pip install memkraft in your current Python environment.