OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Okf Agent Memory

Okf Agent Memory is a Go-based library and CLI for persistent AI-agent project knowledge. It stores knowledge in Git-native files using Open Knowledge Format (OKF) v0.2, provides in-memory BM25 search, and includes an embedded MCP server for agent access. The repository describes a workflow that separates compact working instructions from knowledge retrieved as needed, and provides commands to validate bundles, searc

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Okf Agent Memory — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you are building an agent workflow around persistent project knowledge and want that knowledge stored in Git-native files, retrieved with local BM25 search, or made available through the documented MCP server. The CLI also supports bundle validation and concept maintenance.

Who it is for

  • Developers building AI coding agents that need persistent project context across sessions.
  • Teams maintaining project decisions, constraints, and operational knowledge as version-controlled files.
  • Agent application builders who want a local BM25 retrieval component or an MCP-accessible knowledge bundle.

Documented uses

  • Keep project facts and decisions in a persistent, Git-native knowledge bundle.
  • Search for relevant project concepts and inspect their relationships when an agent needs context.
  • Check constraints associated with a source file before editing it.
  • Validate a bundle and create or update knowledge concepts from the CLI.
  • Provide an MCP server configuration that gives an agent access to a project knowledge directory.
Before you choose

Requirements & limits

  • The tool is a library and CLI for developers to integrate; it is not itself an AI model or hosted agent.
  • The documented search uses in-memory BM25 scoring, which differs from vector search.
  • The repository describes OKF v0.2 knowledge bundles and progressive disclosure through hierarchical indexes and links.
  • MCP access uses the embedded server; the README example invokes the CLI with mcp and a knowledge-directory path.
  • The repository identifies Go as its implementation language and MIT as its license.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Build the CLI and search a knowledge bundle

Compile the standalone okf executable and use its documented validation, search and concept-inspection commands.

Before you start

  • A clone of the repository
  • Go 1.26, as identified by the repository tooling badge
  • A project knowledge bundle

Steps

  1. Run the command below from the repository; the quickstart says this generates bin/okf.

    Documented command
    make build

  2. Validate a bundle with ./bin/okf validate knowledge --strict --drift.

  3. Search it with ./bin/okf search "architecture layers" knowledge.

  4. Inspect a concept with ./bin/okf show architecture/layers knowledge --json.

Expected outcome

A built CLI that can validate a knowledge bundle, retrieve matching concepts, and display a concept with JSON output.

Guide limitations

  • The documented commands describe local CLI use; they do not establish compatibility with a particular agent or project.
  • The search is lexical BM25 retrieval, not vector or embedding-based search.
Documentation 1 ↗
Source-based practical guide

Configure the CLI as an MCP server

Point an MCP client configuration at the CLI and a project knowledge directory.

Before you start

  • A built okf executable
  • A project knowledge directory
  • An MCP client that accepts the documented server configuration format

Steps

  1. Add an mcpServers entry named okf-memory to the client configuration.

  2. Set command to the path of the built okf executable.

  3. Set args to mcp and the path of the project’s knowledge directory.

Expected outcome

The example configuration specifies the executable and knowledge directory for the embedded MCP server.

Guide limitations

  • The README gives an example configuration for Claude Desktop or Cursor; it does not document client-specific setup beyond that example.
  • The configuration example is not evidence of a compatibility guarantee.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
okf-memory/okf-agent-memory · Official homepage title · retained in its source language
How we review sourcesSuggest a correction