Open Source Model Open Source

Cog

Cog provides a centralized, plain-text memory layer for AI agents, allowing them to share and maintain context across projects and tools.

3/10 MegaOne Score

Visit Cog →

Price Free / Open Source

Category Open Source Model

Model Open Source

Company Marcio Puga

Team Size 50

Total Funding $98M

About Cog

Cog is a shared memory system for AI agents, designed to overcome the issue of siloed memory across different tools and projects. It uses structured plain-text files (markdown) that any agent can read, search, and maintain. The system incorporates three primitives: L0 headers for progressive context loading, three tiers of memory (hot, warm, glacier), and a single source of truth with wiki-link cross-references. It operates without a server, database, or application code, relying solely on conventions, and can optionally run a pipeline for automated maintenance like archiving and consolidating data.

Why We Scored It 3/10

3/10

Cog appears to be a small personal project (1 employee, 366 GitHub stars) with limited market impact and no clear evidence of significant adoption or industry relevance in 2025-2026. While it provides interesting functionality as a Claude Code extension, its niche position and lack of substantial user base or funding indicate poor market performance.

Features & Platform

Free Tier

Cog Coverage