Getting Started

This guide takes you from a new account to a working cache with knowledge, rules, and at least one connected assistant. Each step below takes a few minutes and builds on the one before it.

1. Sign in and open the dashboard

Create an account or sign in through the product's authentication flow. Accounts belong to a team, which owns caches and billing. Onboarding may prompt you to name a team or create a first cache before you reach the dashboard.

Outcome: You are signed in and viewing the dashboard.

2. Create your first cache

A cache is an isolated memory space containing chunks, databases, an inbox, and its own settings. Click New Cache in the sidebar (/dashboard/caches/new), choose a template or start blank, and give the cache a name.

Your plan sets the number of caches a team can create. For background, read the Knowledge Layer for caches and chunks, and Structure for how databases and links fit together.

Outcome: A named cache exists. Select it in the sidebar to make it the active context for chat and workflows.

3. Learn the dashboard layout

The left rail is split into Workspace (team-wide tools such as rules, blueprints, plugins, and activity) and Caches (each memory space and its structure). Under a selected cache, you will see Settings and Structure (Inbox, databases, and documents).

For a labeled tour of every dashboard area, see the Dashboard overview.

Outcome: You can locate Workspace tools, Cache settings, and the Structure tree without hunting.

4. Add your first knowledge

Content lives in chunks. Each chunk has a title, body text, optional structured fields, and an embedding for semantic search. To start:

  • Drop rough notes into Inbox, then file them into a database or document when ready.
  • Create rows in a database (Tasks, People, Decisions, Meeting Notes, etc.) when you want typed fields and filters.
  • Run a search in the cache to confirm the assistant can retrieve what you stored.

To push content from scripts or other systems, use scoped Hook API keys created from cache settings.

Outcome: Your cache contains at least one chunk that search can return.

5. Set rules and per-cache instructions

Workspace System Rules (/dashboard/rules) apply to the product's own in-app chat on web and mobile. Use them for global tone and guardrails across every conversation.

Each cache also has system instructions in its settings. These shape AI behavior whenever that cache is the active context. See System Rules for how global context composes with blueprints and data bindings at runtime.

Outcome: Your assistant follows the standing instructions you want applied across every interaction with this cache.

6. Connect an AI assistant

To use ContextCache from Claude, Cursor, or Claude Code, add the MCP server to that client using the connection details from the product and approve access for the cache you want to share. External clients have their own configuration and do not inherit Workspace System Rules.

Start here: MCP Server (overview and client-specific guides).

Outcome: An external assistant can list your caches and read their contents on your behalf.

7. Run a blueprint and try chat

Blueprints are versioned, configurable workflows. Browse the built-in catalog under /dashboard/blueprints and run one from chat, on a schedule, or from an integration, depending on how it is configured.

Chat in the dashboard runs against the active cache. Use it to search, ask questions, and invoke blueprints or agents inline. For detail, see Automations, Blueprints, and Chat.

Outcome: You have run one workflow end-to-end and used chat to retrieve stored content.

What to read next

Getting Started · Documentation · ContextCache