okf-agent-memory

OKF Agent Memory

A Domain-Neutral, Git-Native Persistent Project Memory for AI Agents based on the Open Knowledge Format (OKF) v0.2.

---

🌟 Overview

Conversations with AI agents reset when context windows close. Valuable architectural decisions, domain discoveries, and operational facts are lost unless stored persistently.

Traditional approaches suffer from two fatal failure modes:

1. The Prompt Monolith: Stuffing all domain knowledge and rules into AGENTS.md or CLAUDE.md creates massive context bloat and causes attention drift (agents ignore critical instructions).
2. The RAG Blindspot: Dumping behavioral rules into vector databases fails because agents never semantically search for operational constraints (e.g. formatting or security rules) during general tasks.

OKF Agent Memory resolves this dilemma with the Dual-Memory Agent Architecture (DMAA):

flowchart TD
    subgraph PUSH["1. Normative Working Memory (Push Layer)"]
        direction TB
        C1["Canonical AGENTS.md (capped at 400 tokens)"]
        C2["Domain Codex (Invariants, Ethics, Tone)"]
        C3["OKF Memory Bridge (Deterministic Triggers)"]
        C4["Agent Action Grammar (AAG) Micro-Syntax"]
    end

    subgraph PULL["2. Semantic Domain Memory (Pull Layer)"]
        direction TB
        O1["OKF v0.2 Knowledge Bundle (knowledge/)"]
        O2["0 Tokens baseline in system prompt"]
        O3["Selective Retrieval via okf_search / okf_show"]
        O4["Persistent Graph of Decisions, Facts & Runbooks"]
    end

    INPUT["User Request"] --> PUSH
    PUSH -->|Enforces Domain Codex & Triggers| AGENT["AI Agent (LLM)"]
    AGENT -->|Selective Retrieval| PULL
    PULL -->|Context & Facts| AGENT
    AGENT --> OUTPUT["Deterministic Response"]

The Universal Composition Model

In DMAA, every agent configuration is structured by a universal composition:

$$\text{AGENTS.md} = \underbrace{\text{Domain Codex (AAG)}}{\text{Project Invariants, Tone, Guardrails}} + \underbrace{\text{OKF Memory Bridge}}{\text{Standardized Triggers: Search-Before-Write}}$$

• Layer 1: Normative Working Memory (Push Layer): A permanent, ultra-compact behavioral codex (capped at 400 tokens) expressed in Agent Action Grammar (AAG). Loaded at session start, enforcing zero-tolerance invariants.
• Layer 2: Semantic Domain Memory (Pull Layer): An OKF v0.2 knowledge bundle (knowledge/) that consumes 0 tokens at baseline and is queried on-demand in microseconds.

flowchart TD
    L1["1. OKF v0.2 Specification<br/>(Normative Markdown & YAML Format)"]
    L2["2. Agent Memory Convention & DMAA<br/>(Dual-Memory Model, Search-Before-Write, Trust)"]
    L3["3. Agent Skill & AAG Codex<br/>(Agent Action Grammar, Workflows, Triggers)"]
    L4["4. Tooling Layer: Go Library & CLI<br/>(Deterministic Parsing, Validation, BM25, MCP)"]
    L5["5. Project Knowledge Corpus<br/>(knowledge/ OKF Bundle)"]

    L1 --> L2
    L2 --> L3
    L3 --> L4
    L4 --> L5

---

⚡ Key Highlights

• Dual-Memory Cognitive Architecture (DMAA): Separates normative push working memory (AGENTS.md codex) from semantic pull domain memory (knowledge/ bundle), completely eliminating prompt bloat.
• Agent Action Grammar (AAG): Ultra-compact, deterministic ASCII micro-syntax saving ~78–85% tokens compared to natural language prompt instructions.
• Blazing Fast Performance (<300µs Search, ~4ms Graph Validation): In-memory BM25 retrieval and bundle validation execute in microseconds without VM spin-up or network roundtrips.
• 100% Git-Native & Zero Vendor Lock-in: Everything is version-controlled plain text. Inspect, audit, and review your agent's memory using standard git diff and git log. No external database required.
• Zero API Costs for Memory Retrieval: Local lexical BM25 indexing eliminates recurring vector embedding API costs and network roundtrips.
• Built on Google OKF v0.2: Uses the open standard format for agent knowledge with full support for provenance (sources), trust tiers (generated vs. verified), and lifecycle metadata (status, stale_after).
• Solves Context Bloat & Memory Rot: Employs Progressive Disclosure (hierarchical index.md files and link graphs) so agents only load the exact concepts they need.
• Search-Before-Write Principle: Mandates querying existing memory before authoring, preventing concept duplication and hallucinated divergence.
• Governance & Code-to-Knowledge Binding: Bind architecture decisions directly to source files via code_refs and query active constraints/holds via --for-path before modifying code.
• Truly Domain-Neutral: Designed for Software Engineering, Coaching, Scientific Research, Literature Reviews, and Operations.

---

📊 Performance Benchmarks

Built in Go with zero external dependencies, okf is engineered for high-frequency agent tool calling loops:

Benchmark Metric	Python / Vector DB Runtimes (Mem0, Letta)	Deno / Node.js Tooling	OKF Agent Memory (Go)
Concept Search Latency	150ms – 800ms (Embedding API + Vector DB)	40ms – 120ms	< 300 µs (Microseconds, In-Memory BM25)
Full Corpus Parse & Graph Validation	200ms – 1.5s	80ms – 250ms	~4.0 ms (50+ concepts, bidirectional graph)
Process Cold-Start Overhead	250ms – 600ms (Python VM boot)	80ms – 180ms (V8 / Deno boot)	< 4 ms (Compiled Single Binary)
Retrieval Cost per 1,000 Queries	~$0.10 – $0.50 (Embedding tokens)	$0.00	$0.00 (Zero API cost, fully local)
Memory Footprint (RSS)	~120 MB – 350 MB	~60 MB – 140 MB	< 15 MB

[!TIP] Reproduce Locally with your own LLM: We provide an automated benchmark runner in pure Go to verify Time-To-First-Token (TTFT) speedups and -80% token reduction on your local hardware (LM Studio / Ollama with Gemma, Qwen, Llama). Run make benchmark or explore the Progressive Disclosure Benchmark Suite.

---

🚀 Quickstart

1. Build the Tooling

Clone the repository and compile the standalone okf executable:

make build

This generates the standalone binary at bin/okf.

2. Basic CLI Commands

# Validate bundle conformance, graph connectivity, and description drift
./bin/okf validate knowledge --strict --drift

# Evaluate bundle quality scorecard (3 pillars: maintenance, stewardship, integrity)
./bin/okf validate knowledge --scorecard

# Gate bundle quality in CI or pre-commit pipelines (fails if score < 80)
./bin/okf validate knowledge --min-score 80

# Search concepts via in-memory BM25 scoring
./bin/okf search "architecture layers" knowledge

# Discover constraints and active holds governing a specific source file before editing
./bin/okf search --for-path pkg/okf/types.go knowledge

# Inspect a concept and its relationships (with --json support)
./bin/okf show architecture/layers knowledge --json

# Create a new concept with automated log.md and index.md bookkeeping
./bin/okf create decisions/auth-flow knowledge \
  --type Decision \
  --title "OAuth2 Authorization Flow" \
  --desc "Standardized on PKCE for client authentication."

# Update an existing concept
./bin/okf update decisions/auth-flow knowledge \
  --desc "Updated OAuth2 PKCE token refresh interval."

# Bootstrap full agent memory stack into any target project
./bin/okf bootstrap /path/to/project --name "My Project"

# Initialize only a bare OKF bundle in any directory
./bin/okf init my-project/knowledge

# Zero-Knowledge Sync: initialize vault and print Emergency Kit
./bin/okf hub init-vault knowledge

# Zero-Knowledge Sync: push or sync changes with the Hub (optional: --token or OKF_HUB_TOKEN)
./bin/okf hub push knowledge --password "pass" --secret-key "XXXX-..." --auth-token "my-token"
./bin/okf hub sync knowledge --password "pass" --secret-key "XXXX-..."

# Pull external knowledge bundles from OKF Registry (registry.okf-memory.dev) or Git
./bin/okf pull nextjs-15
./bin/okf pull nextjs-15 --min-score 80   # Gate bundle quality before installation
./bin/okf pull peter/django-5-rules
./bin/okf pull github.com/acme/[email protected]

# Restore all locked vendor bundles in fresh environments
./bin/okf restore
./bin/okf restore --min-score 80          # Gate all restored bundles

# Inspect and manage installed vendor bundles
./bin/okf vendor list
./bin/okf vendor remove nextjs-15

3. Multi-Scope Memory Layering & Scoped Resolution

OKF Agent Memory organises knowledge across four deterministic memory scopes. Higher layers strictly shadow identical concept IDs in lower layers during search, ensuring local project decisions always take precedence over external upstream packages or machine baselines:

flowchart TD
    subgraph Scopes ["OKF 4-Tier Memory Hierarchy"]
        P["<b>Project</b> (Priority 100)<br/>Prefix: <i>none</i> (e.g. decisions/auth)<br/>Location: ./knowledge/"]
        V["<b>Vendor</b> (Priority 70)<br/>Prefix: @bundle/... (e.g. @nextjs-15/routing)<br/>Location: .okf/vendor/"]
        U["<b>User</b> (Priority 50)<br/>Prefix: user:... (e.g. user:guidelines/style)<br/>Location: ~/.okf/"]
        S["<b>System</b> (Priority 10)<br/>Prefix: system:... (e.g. system:corp/policy)<br/>Location: /etc/okf/"]
    end

    P -->|shadows| V
    V -->|shadows| U
    U -->|shadows| S

Scope Specification Matrix

Scope	Link / Reference Syntax	Canonical URN	Storage Location	Priority	Precedence & Rules
project	decisions/auth.md	(bundle relative)	./knowledge/	100	Local Project Memory. Authoritative SSoT for current repository; strictly shadows identical IDs from vendor, user, and system layers.
vendor	@nextjs-15/routing.md@peter/django-rules/auth.md	okf://@nextjs-15/routingokf://@peter/django-rules/auth	.okf/vendor/<bundle>/	70	External Packages. Pinned dependencies pulled from OKF Registry (registry.okf-memory.dev) or Git. Shadows user and system.
user	user:guidelines/style.md	okf://user/guidelines/style	~/.okf/	50	Personal Agent Memory. Developer preferences and cross-project notes. Shadows system.
system	system:corp/policies.md	okf://system/corp/policies	/etc/okf/	10	Enterprise / Machine Memory. Infrastructure baselines and compliance policies.

Filtering by Scope in Search & Show

# Default (--scope all): Search across all layers with automatic shadowing & priority ranking
./bin/okf search "authentication"

# Search strictly within vendor packages
./bin/okf search "routing" --scope vendor

# Search strictly within local project memory
./bin/okf search "architecture" --scope project

# Search user or system layers
./bin/okf search "style" --scope user
./bin/okf search "compliance" --scope system

# Inspect concepts across scopes
./bin/okf show decisions/auth                       # Local project concept
./bin/okf show @nextjs-15/decisions/routing         # Vendor package concept
./bin/okf show user:guidelines/style                # Personal user concept
./bin/okf show system:corp/policies                 # System compliance concept

Bundle Scope Isolation (scopes in knowledge/index.md)

By default, all four scopes (project, vendor, user, system) are active. To isolate a project (e.g. preventing agents in a client or open-source repository from accessing personal ~/.okf/ notes), declare an explicit scopes list in the root knowledge/index.md frontmatter:

---
okf_version: "0.2"
scopes: [project, vendor]
---

• Only the listed scopes are searched under --scope all or bare-ID okf show fallback.
• Explicit queries targeting a disabled scope (e.g. okf search --scope user or okf show user:secret) are rejected with an error.
• Empty (scopes: []) or non-canonical scope names fail bundle loading and validation immediately (fail-closed).

4. Bundle Quality Scorecards (OpenSSF-Style)

Evaluate knowledge bundles across three weighted pillars to eliminate "reproducibly wrong" context without maintenance chores:

1. Epistemic Maintenance (35%): Invariant version-bound facts without stale_after remain 100% valid by default; penalties apply only to lapsed review deadlines or deprecated status.
2. Human Stewardship (35%): Strictly distinguishes verified human sign-offs (verified.by: human/* or user/*) from unreviewed synthetic bot dumps. Non-human verifications earn 0 stewardship points.
3. Graph Integrity (30%): Strict OKF v0.2 graph health (0 broken links, 0 orphans, valid DAG).

# Human-readable scorecard breakdown with Grade (A-F)
./bin/okf validate knowledge --scorecard

# Machine-readable JSON output
./bin/okf validate knowledge --scorecard --json

# Quality gating in CI (fails with exit code 1 if score < 80)
./bin/okf validate knowledge --min-score 80

# Pre-pull quality gating (aborts installation & rolls back if score < 80)
./bin/okf pull nextjs-15 --min-score 80

5. Bootstrapping Agent Memory in Any Project

Scaffold the complete OKF Agent Memory architecture into any new or existing repository with a single command:

# Bootstrap full memory stack into target project
./bin/okf bootstrap /path/to/my-project --name "My Service"

This automatically sets up:

• knowledge/ — OKF v0.2 compliant persistent memory bundle (index.md, log.md)
• .agents/skills/okf-memory/ — Embedded agent skill definition and capability guides
• AGENTS.md — Project-tailored operating instructions for AI coding agents
• Makefile — Convenience tasks for validation (make validate) and search (make search q="...")

6. Running as an MCP Server

okf ships with a native Model Context Protocol (MCP) server over stdio to seamlessly connect with Claude Code, Cursor, Codex, and other agent platforms:

./bin/okf mcp knowledge

Example MCP Configuration (claude_desktop_config.json or Cursor):

{
  "mcpServers": {
    "okf-memory": {
      "command": "/path/to/okf-agent-memory/bin/okf",
      "args": ["mcp", "/path/to/project/knowledge"]
    }
  }
}

---

📂 Repository Structure

okf-agent-memory/
├── .agents/                # Active agent skills and agent configuration
│   └── skills/okf-memory/  # Authoritative OKF memory skill for AI agents (Single Source of Truth)
├── benchmarks/             # Progressive disclosure benchmark suite & hardware test data
│   ├── data/               # Monolith docs vs OKF bundle test fixtures
│   └── results/            # Reproducible benchmark logs across 8+ local & cloud LLMs
├── cmd/
│   ├── okf/                # Standalone CLI and embedded MCP server (`stdio`)
│   └── okf-benchmark/      # Automated benchmark runner for LLM TTFT & token measurements
├── docs/                   # Guides, specifications, architecture & release playbook
│   ├── README.md           # Central documentation index & navigation
│   ├── guides/             # User guides, CLI/MCP reference & AI instruction best practices
│   ├── spec/               # OKF convention v0.1, compatibility analysis & architecture RFCs
│   ├── security/           # Data governance, secret prevention & adversarial security audits
│   ├── project/            # Project roadmap, release playbook & multi-agent testing
│   └── releases/           # Versioned release notes & changelog archive (v0.1.0 - v0.8.0)
├── examples/               # Domain-neutral reference DMAA projects (AGENTS.md + OKF v0.2 knowledge/)
│   ├── books/              # Literature & editorial analysis repository
│   ├── coaching/           # Executive coaching & client session repository
│   └── software/           # Microservices architecture & ADR engineering repository
├── knowledge/              # Project's own OKF v0.2 persistent memory bundle
│   ├── index.md            # Root progressive disclosure index (okf_version: "0.2")
│   ├── log.md              # Dated change log (ISO 8601 YYYY-MM-DD)
│   ├── project/            # Overview & value propositions
│   ├── architecture/       # 5-tier architecture, governance model & decisions
│   ├── convention/         # Principles & lifecycle workflows
│   └── roadmap/            # Milestones
├── packaging/              # Distribution packaging
│   └── homebrew/           # Official Homebrew formula & tap instructions
├── pkg/
│   ├── lock/               # Zero-dependency okf.lock parser & serializer
│   ├── okf/                # Core library (parser, validator, BM25, MCP, bootstrap)
│   │   └── assets/         # Embedded bootstrap templates & skills mirrored via `make sync-assets`
│   ├── registry/           # Decentralized OKF Registry client & archive unpacker
│   ├── sync/               # Blind sync engine, 3-way reconcile & hub client/server
│   └── vault/              # AES-256-GCM envelope crypto, Argon2id KDF & CAS blind storage
├── AGENTS.md               # Operating instructions for AI coding agents
├── CONTRIBUTING.md         # Contribution guidelines & development workflow
├── CONTRIBUTORS.md         # Community contributors & acknowledgements
├── CODE_OF_CONDUCT.md      # Contributor Covenant v2.1 code of conduct
├── Makefile                # Build, test, lint, validation & release targets
├── LICENSE                 # MIT License
├── README.md               # Main repository documentation
└── SECURITY.md             # Security policy & reporting guidelines

---

🧪 Testing & Verification

Run the full test suite and validate the repository's self-documenting knowledge bundle:

make check

---

📖 Further Documentation

• Documentation Index — Central directory of all project documentation.
• Getting Started Guide — Comprehensive onboarding guide for agents and humans.
• CLI & MCP Reference — Complete command-line and protocol tools reference.
• Dual-Memory Agent Architecture RFC — Cognitive 2-layer agent memory model (Push codex + Pull knowledge).
• Agent Action Grammar RFC — Deterministic, token-efficient AAG micro-syntax specification for AGENTS.md.
• Agent Instruction Best Practices — Guide to deterministic instruction design and token optimization.
• OKF Agent Memory Convention v0.1 — Behavioral rules and lifecycle specification.
• Contributing Guide — Development setup, quality gates, and pull request standards.
• Security & Privacy Guidelines — Data governance, secret prevention, and PII protection rules.
• Multi-Agent Testing & Evaluation — Test scenarios, compatibility matrix, and benchmarks.
• Project Roadmap & Milestones — Phased development plan.
• Release Playbook — Versioning, CI/CD pipeline, and distribution procedures.
• Release Notes & History — Versioned changelogs and historical release notes archive.
• OKF v0.2 Compatibility Matrix — Specification validation analysis.
• Why OKF Agent Memory? — Detailed value proposition & differentiators.
• Alternatives & Ecosystem Comparison — Comparison with Mem0, Letta, and ad-hoc markdown files.

---

👥 Contributors

Thank you to all the wonderful contributors who have helped build and refine OKF Agent Memory!

Contributions of all kinds are warmly welcomed! See CONTRIBUTING.md and CONTRIBUTORS.md for details.

---

💖 Support & Sponsoring

If you find OKF Agent Memory valuable for your autonomous agent workflows, consider sponsoring the project on GitHub to help support continuous development, security hardening, and spec compliance!

---

📄 License

MIT License. See LICENSE for details.