Skip to content

v0.1.0 — Initial Release: The Git-Native Memory Engine for AI Agents

Latest

Choose a tag to compare

@sknr sknr released this 05 Sep 22:08
· 1 commit to main since this release

OKF Agent Memory v0.1.0

We are excited to announce the initial release of OKF Agent Memory (v0.1.0) — an ultra-lightweight, Git-native, domain-neutral persistent project memory engine for AI coding agents (Claude Code, Cursor, Windsurf, Codex, and local LLMs), built on the Google Open Knowledge Format (OKF) v0.2 specification.

🌟 Key Highlights

  • ⚡ Sub-300µs BM25 Search: Zero vector database dependencies. Search your repository's knowledge graph directly in memory in microseconds with 0.00 € API costs.
  • 🔌 Built-In Stdio MCP Server: Native Model Context Protocol server inside the single binary (okf mcp knowledge). Plug and play with Claude Desktop, Cursor, and IDEs.
  • 🛡️ 100% Git-Native & Auditable: All memory lives as plain Markdown files with YAML frontmatter in knowledge/. Inspect changes via git diff and standard PR reviews.
  • 🗂️ Progressive Disclosure: Slashes context window bloat by up to 94%. Agents dynamically retrieve 300-token atomic concepts instead of dumping 20k-token monolith files.
  • 🔒 Trust & Provenance Tiers: Formal distinction between authoritative human decisions (verified: human:lead@...) and agent drafts (generated: agent/...).
  • 🚀 1-Command Bootstrapping: okf bootstrap . scaffolds memory, installs agent skills, and non-destructively enriches AGENTS.md.

📦 Installation

# macOS / Linux via Homebrew:
brew install okf-memory/tap/okf

# Universal cURL 1-Liner:
curl -fsSL https://okf-memory.dev/install.sh | sh

# Go Toolchain:
go install github.com/okf-memory/okf-agent-memory/cmd/okf@v0.1.0