# Shared context for your team and your agents.

Bring your documents, conversations, and code together in one knowledge graph. Find the connections, ask better questions, and give your AI agents the context to keep moving.

## A shared understanding of your work

People, projects, decisions, meetings, and ideas rarely live in one place. OpenContext turns the material you already have into connected notes, each linked back to its sources. Explore the graph to see the bigger picture, then open any connection to read what supports it.

Connect documents and spreadsheets, Google Drive, Gmail, GitHub and GitLab repositories, bookmarks, or exported conversations. You can also start with a single file, link, or pasted thought. [See how your knowledge comes together](https://getopencontext.com/how-it-works.md).

## Ask a question. Follow the evidence.

Chat with your knowledge, revisit a conversation, and open the notes behind an answer. Insights surface connections across your work. Evidence, history, and review keep the underlying knowledge inspectable, so a useful answer is also one you can check.

Your workspace brings these views together: a searchable library, an expansive graph, conversations, project knowledge, and a record of changes. You can move from the big picture to one source without losing your place.

## Give your agents a memory

The context you build is useful beyond this app. Coding agents and chat assistants can read it through a command-line tool or a Model Context Protocol (MCP) server. Linked repositories bring the relevant people, architecture, and decisions into an agent's work, with sources attached.

Team activity connects coding sessions with the repositories and files they touch. Review that activity alongside the knowledge your team shares. [Explore the agent connection](https://getopencontext.com/for-agents.md).

## Your knowledge stays yours

The graph is a folder of plain markdown, with links and source references. You can read it in your own editor, keep its history in git, download it, or move it somewhere else.

OpenContext stores your files on the computer running the app, with no publisher collection service, telemetry, or analytics. Local rules can hold sensitive identifiers before distillation and redact matched values from generated notes. Enabled AI features send content to your configured provider, under your account's terms. [Explore privacy and security](https://getopencontext.com/privacy.md): what the rules detect, what is kept, and what your model receives.

## Start with what you already know

Open your workspace to explore the graph, pick up a conversation, or bring in your next document. For your own installation, the [getting started guide](https://getopencontext.com/install.md) walks through setup and connecting a model. OpenContext is free during the closed beta; your files and the knowledge you create remain yours.

For developers and AI agents, the [agent guide](https://getopencontext.com/for-agents.md), [changelog](https://getopencontext.com/changelog.md), and [source repository](https://github.com/Alec-Merle/OpenContext) describe what ships today. Every page also has a markdown twin, indexed in [llms.txt](https://getopencontext.com/llms.txt).

## Measured results

Three activated agent comparison pairs on one task averaged 20.6% lower cost and 32.7% fewer cache-read tokens. A blind judge preferred the graph arm in one of three pairs. These are early results on one codebase and one model, not a forecast. [Read the full evidence and limitations](https://getopencontext.com/evidence.md).

The interactive homepage includes a fictional example graph, seven ways to arrange its entities, local example answers, supported source types, and an illustrative savings calculator. The demo does not call a model or read your workspace.