</>MCP Agents Market
Agent

gbrain

by garrytan28.6kTypeScriptUpdated 2026-08-17

Garry's Opinionated OpenClaw/Hermes Agent Brain

Claude CodeClaude DesktopCodexCursorWindsurf

GBrain is an AI sub-agent that functions as a persistent memory and knowledge synthesis layer for coding agents and autonomous AI systems. Unlike traditional search tools that return raw pages, it synthesizes answers with citations, performs gap analysis, and maintains a self-wiring knowledge graph that tracks relationships between people, companies, and entities. Created by Y Combinator President Garry Tan, it powers his production deployments handling over 155,000 pages and runs continuously to ingest meetings, emails, and voice calls while enriching institutional memory overnight.

Key Features

Synthesized answers with citations instead of raw search results, including explicit gap analysis showing what the brain doesn't know yet
Self-wiring knowledge graph that extracts entity relationships (works_at, invested_in, founded, advises) with zero LLM calls
Multi-tenant company brain mode with per-user scoping and zero-leak privacy across team members
24/7 autonomous dream cycle that continuously ingests data, enriches entities, fixes citations, and consolidates memory
Local-first PGLite database setup in under 2 seconds with no Docker or server required
Seven-verb memory protocol (recall, remember, entity, synthesize, forget, context_pack, delta) frozen and additive-forever
Bootstrapped personal agent installation via Codex or Claude Code in ~15 minutes with interview-driven identity creation
Benchmark-verified retrieval: 49.1% P@5 and 97.9% R@5, outperforming vector-only RAG by 31.4 points

Use Cases

  • 01Meeting preparation by synthesizing everything known about a person or company with open action items and knowledge gaps
  • 02Persistent memory layer for coding agents across chat sessions, enabling cross-session recall without context loss
  • 03Team institutional memory with privacy-scoped access where each member sees only their authorized data
  • 04Autonomous ingestion of Gmail, calendar, contacts, and offline archives to build a personal knowledge base
  • 05Always-on agent deployments on OpenClaw or Hermes that enrich data and consolidate memory overnight
  • 06Converting Codex or Claude Code into a persistent personal agent with private GitHub repo storage

Related Agents

View more

unsloth

Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more.

Agent72.6k

codegraph

Pre-indexed code knowledge graph, auto syncs on code changes, for Claude Code, Codex, Gemini, Cursor, OpenCode, AntiGravity, Kiro, and Hermes Agent — fewer tokens, fewer tool calls, 100% local

Agent66.6k

claude-code-best-practice

from vibe coding to agentic engineering - practice makes claude perfect

Agent64.6k

system_prompts_leaks

Extracted system prompts from Anthropic - Claude Fable 5, Opus 5, Claude Design, Claude Code. OpenAI - ChatGPT GPT-5.6-Sol, Codex. Google - Gemini 3.5 Flash, 3.1 Pro, Antigravity. xAI - Grok, Cursor, Copilot, VS Code, Perplexity, and more. Updated regularly.

Agent63k

gbrain — FAQ

What is GBrain and how does it differ from regular search?+

GBrain is an AI sub-agent that provides synthesized answers with citations and gap analysis, rather than just returning a list of matching pages. It maintains a knowledge graph of entity relationships and tells you explicitly what it doesn't know yet, making it a true memory layer for autonomous agents.

How do I install GBrain for Codex or Claude Code?+

Open Codex or Claude Code in a new empty folder, then paste the bootstrap instructions from BOOTSTRAP_FOR_AGENTS.md. The agent will interview you, create your identity files, set up a local PGLite brain in about 2 seconds, and create a verified private GitHub repo. The entire setup takes approximately 15 minutes.

Which AI clients and platforms does GBrain work with?+

GBrain supports Claude Code, Codex (ChatGPT desktop), Cursor, Windsurf, Claude Desktop (Cowork), Grok Build, opencode, OpenClaw, and Hermes. It exposes tools via MCP protocol in both stdio and HTTP modes.

Do I need API keys to use GBrain?+

No, GBrain works with zero API keys in keyword search mode using memory your agent writes itself. Optional keys (OpenAI, Voyage, or Anthropic) upgrade capabilities to include semantic search and automatic fact extraction.

Is GBrain free and open source?+

Yes, GBrain is open source and installed directly from the GitHub repository. You own the brain data (stored as markdown files), and there are no subscription fees, though running always-on deployments on OpenClaw or Hermes incurs server and API costs.

How does the company brain multi-tenant mode work?+

Each team member gets a login-scoped slice of the brain and can only see data they're authorized to access. GBrain has been fuzz-tested across all read operations (search, list, lookup, multi-source reads) with zero leaks, making it safe for shared institutional memory.

How do I install gbrain?+

Open the source repository on GitHub and follow its README. gbrain is a agent — MCP Agents Market links you directly to the official repo.

Is gbrain free?+

gbrain is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches

Knowledge & Memory AgentsDeveloper Tools AgentsProductivity & Workflow AgentsBest MCP ServersBest SkillsBest PluginsTop rated listingsClaude Code toolsBrowse all categories