Context management

AI context management, done once.

An AI agent or LLM is only as good as the context it gets. Ensemble keeps yours in one place: people, projects, notes, meetings, files and how you like work done. Then it hands that context to the assistant and to your editors, so you never type it in again.

Context that builds itself from your work

People and repos are learned from the mail, chat, issues and meetings you connect. A person is recognised by any address or handle they use. Link a Slack channel, GitHub repo or Linear project to a project, and what arrives from it lands in the right place.

One context for every AI agent

Ensemble's MCP bridge gives editors and agents the same briefs, notes, tasks, people and skills the assistant uses. Set it up with one command for VS Code, Cursor, Claude Code, Codex, Gemini CLI, Windsurf, Zed and others, or use the hosted MCP address with a read-only key.

Skills: how you like things done

Versioned skills hold the way you write pull request descriptions, commit messages and replies. Edit them like any page; agents and the assistant use them when they draft, so the first draft already sounds like you.

Spaces keep contexts apart

Keep work, study and side projects in separate spaces, each with its own notes, tasks and context. Share a space with the people who need it, and keep your settings, keys and connected apps private.

You decide what leaves

Connectors only read. Secret-looking strings are redacted before what they read is stored, tokens are encrypted, and disconnecting an app can delete what it read. MCP keys are read-only, and a model sees only what you send to the provider you chose.

Context you can audit

Every agent action is written to a hash-chained, insert-only ledger with what it read and who approved it. Every prompt Ensemble uses can be read in Settings, and the code is public on GitHub.

FAQ

Questions, answered.

What is AI context management?

AI context management means collecting, organising and supplying the background an AI model needs to do useful work: who people are, what a project is about, what was decided, and how you like things done. Ensemble keeps that context in one workspace and gives it to the assistant and to AI agents, so you do not paste it into every prompt.

How do I give Claude Code, Cursor or Copilot my context?

Install the Ensemble CLI and run ensemble mcp setup for your editor, or add the hosted MCP address with a key from the Connect page. The editor can then read your briefs, notes, tasks, people and skills from Ensemble, and with a key that allows changes, update your tasks and pages too. AI extensions such as Claude Code inside VS Code have their own setup, which the Connect page shows.

What is MCP?

MCP, the Model Context Protocol, is an open standard that lets AI apps and agents read tools and context from other software. Through Ensemble's MCP server an agent reads your context, and with a key that allows changes it can also update tasks, pages, boards, diagrams and plots; each change can be undone in Ensemble.

Is my context sent to AI providers?

Only when you use a model feature, and only to the provider you picked with your own key. Ensemble does not sell your data or use it for advertising. The privacy policy at ensemblework.com/privacy has the details.

Keep reading

More of the workspace.

Bring your agent into the room.

Create a free account, save your notes and context, and bring your own model key when you are ready to use an agent.