# gcontext

> State management for AI agents

gcontext is an MCP server that serves a folder of plain text files as your
agent's persistent state. Connections, modules, scripts, and decisions
accumulate over time. Your runtime does the reasoning, the folder remembers.

## Docs

- [Get started](https://gcontext.ai/#get-started): Install gcontext and connect an MCP client
- [Connections](https://gcontext.ai/connections): How agents reach external services
- [Agents](https://gcontext.ai/agents): Pre-built workflows you install into your instance
- [Browse agents](https://gcontext.ai/agents/browse): Search the agent registry
- [Changelog](https://gcontext.ai/changelog): Release history

## Developer resources

- [OpenAPI spec](https://gcontext.ai/openapi.json): Machine-readable API description (OpenAPI 3.1)
- [llms.txt](https://gcontext.ai/llms.txt): AI-readable site navigation following the llms.txt convention

## Concepts

- [Agent state](https://gcontext.ai/concepts/agent-state): Definition and management patterns for durable agent knowledge

## Compare

- [LangGraph memory](https://gcontext.ai/compare/langgraph-memory): Checkpointers, the Store API, and a file layer for cross-tool knowledge
- [Claude Skills](https://gcontext.ai/compare/claude-skills): Instructions vs memory, and how the two layers work together
- [Claude subagents](https://gcontext.ai/compare/claude-subagents): Sharing context between Claude Code subagents
- [Vercel Eve](https://gcontext.ai/compare/vercel-eve): Eve handles agent execution, gcontext handles persistent state
- [CrewAI memory](https://gcontext.ai/compare/crewai-memory): How CrewAI handles memory and where a plain-file state layer fills the gaps

## Memory

- [AI agent memory](https://gcontext.ai/memory/ai-agents): Four approaches to persistent agent memory compared
- [Claude Code memory](https://gcontext.ai/memory/claude-code): What persists and what does not in Claude Code

## Guides

- [Agent registry](https://gcontext.ai/guides/agent-registry): Install, update, and remove pre-built agents with one command
- [Stateful MCP server](https://gcontext.ai/guides/build-mcp-server-with-state): Build an MCP server with persistent state
- [Share context across tools](https://gcontext.ai/guides/share-context-claude-code-cursor): Connect multiple AI clients to one state folder

## Source

- [GitHub](https://github.com/bleak-ai/gcontext): Source code (MIT license)
- [PyPI](https://pypi.org/project/gcontext-ai/): Python package

## Quick start

    uv tool install gcontext-ai
    gcontext init my_instance
    gcontext up my_instance