Getting Started
Understanding ContextStream.
Whether you're a developer or managing projects without code, here's how ContextStream organizes your AI's memory and context.
Architecture
How it all connects.
Layer 01 · Workspace
Your team or personal space
“Acme Corp” or “My Personal Projects”
Layer 02 · Project
Project: Backend API
Code repository
Layer 02 · Project
Project: Marketing
Documents & plans
Within each project, AI remembers
Memory
Decisions, conversations, learnings that persist forever.
Context
Indexed files, docs, and searchable content.
Knowledge
Connected insights showing how things relate.
Workspace = your container
Think of a workspace like a company or team folder. Everything inside belongs together.
Examples
- •“Your Company” or “My Personal Projects”
- •“Client: Acme” — a client's work
Project = specific work
Each project focuses on one thing. Could be a codebase, a document set, or an initiative.
Examples
- •A GitHub repository
- •Marketing campaign docs
- •Research notes collection
FAQ
Frequently asked questions.
Common questions when getting started.
Group 01
For developers
How does ContextStream integrate with my IDE?
ContextStream uses the Model Context Protocol (MCP) to integrate directly with AI-powered IDEs like Cursor and Claude Code. Once configured, your AI assistant automatically has access to your project's memory and context. Just add our MCP server to your IDE config and you're set.
What happens when I index my codebase?
When you index a repository, ContextStream analyzes your code structure, extracts semantic meaning, and creates searchable embeddings. This means your AI can find relevant code by describing what you need, not just by filename or exact text matches. We also detect function relationships, dependencies, and code patterns.
How do I organize multiple repositories?
Create one workspace for related projects (e.g., “Company Backend”) and add each repository as a separate project within that workspace. The AI can then understand relationships across repos while keeping context organized. For unrelated work, use separate workspaces.
What's stored in memory vs. indexed content?
Indexed content = your actual files, code, and documents. This is searchable context.
Memory = conversations, decisions, preferences, and learnings. This persists and grows over time.
Think of indexed content as “what exists” and memory as “what happened and what we learned.”
Memory = conversations, decisions, preferences, and learnings. This persists and grows over time.
Think of indexed content as “what exists” and memory as “what happened and what we learned.”
Can I use ContextStream with private/enterprise repos?
Yes! We support GitHub, GitLab, Bitbucket, and local repositories. Your code is encrypted in transit and at rest. The MCP server skips common dependency/build directories and large files by default, and only indexes common code/config/doc file types. For complete data sovereignty, self-hosting is available on Enterprise plans.
How does the knowledge graph help with code?
The knowledge graph maps relationships: which functions call which, how modules depend on each other, and connections between decisions and code. Ask things like “what would break if I change this function?” or “show me all code related to authentication.” It's like having x-ray vision into your codebase.
Group 02
For teams & non-coders
I don't code. How can ContextStream help me?
ContextStream works with any content, not just code! Use it to give your AI memory about:
- •Marketing strategies and brand guidelines
- •Meeting notes and decisions
- •Research and documentation
- •Project plans and requirements
Your AI assistant will remember context from past conversations and reference relevant documents automatically.
How do team members share context?
Workspaces are shareable! Invite team members to your workspace and everyone's AI assistants will have access to the same memory and context. When someone makes a decision or adds important information, it's available to the whole team. No more “I didn't know we decided that!”
What's a good way to organize my workspace?
Start simple:
- •One workspace per team, company, or major initiative
- •One project per distinct focus area (product, campaign, research topic)
- •Let memory grow naturally — just use your AI and it captures important context
You can always reorganize later. The key is to start using it!
How is this different from just saving chat histories?
Chat histories are sequential and hard to search. ContextStream:
- •Extracts key insights automatically from conversations
- •Connects related ideas across different sessions
- •Searches by meaning, not just keywords
- •Provides context proactively when relevant
It's the difference between a pile of notes and a smart assistant who remembers everything important.
Is my team's data private and secure?
Absolutely. Your data is encrypted at rest and in transit. Each workspace is isolated. We don't train on your data or share it. Enterprise plans include additional controls like SSO, audit logs, and the option to self-host for complete data sovereignty.
What documents and file types are supported?
We support most common formats:
- •Documents: PDF, Word, Google Docs, Markdown
- •Spreadsheets: Excel, CSV, Google Sheets
- •Text: Plain text, JSON, YAML
- •Code: All major programming languages
Upload files directly or connect to your existing document repositories.
Group 03
Token saving & costs
How does ContextStream save AI tokens?
Traditional AI chat often includes your entire conversation history with every prompt, so prompts grow as conversations get longer. ContextStream stores context externally and provides tools like
context_smart that retrieve only relevant context on-demand. This keeps prompts bounded and can significantly reduce token usage while maintaining context quality.What is context_smart and how does it work?
context_smart is the key token-saving tool. When called with a user message, it:- 1.Analyzes the message to understand what context is needed
- 2.Searches decisions, memory, and code for relevant information
- 3.Returns a minified format:
D:Decision|M:Memory|P:Preference
The AI can include this compact context instead of resending your full chat history.
How do I enable token-saving in my editor?
Add editor rules (e.g.,
.windsurf/rules/contextstream.md or .cursorrules) that instruct the AI to call context_smart before every response. You can also use generate_rules to auto-generate these files. See our MCP Integration Guide for details.What about long conversations - do they still get expensive?
No! With ContextStream, use
session_compress to extract key decisions, preferences, and insights from your chat history and store them as memory events. Then you can start fresh conversations without losing context. Each new conversation just calls session_summary or context_smart to get the relevant context.Group 04
General questions
How do I get started?
- 1.Sign up for Pro
- 2.Create a workspace for your team or project
- 3.Add the MCP server to your AI tool (Claude, Cursor, etc.)
- 4.Start using AI — context is captured automatically
The whole setup takes about 5 minutes. Check out our Quickstart Guide for detailed steps.
What's included in the Pro plan?
Pro gives you full access to all features:
- •3 workspaces
- •Unlimited projects
- •25,000 operations per month
- •All search types (semantic, hybrid, pattern)
- •Graph-Lite knowledge graphs (module-level dependencies + 1-hop impact)
- •MCP integration
Pro is $19/mo.
Which AI tools work with ContextStream?
Any tool that supports the Model Context Protocol (MCP):
- •Claude Code — Anthropic's coding assistant
- •Cursor — AI-powered code editor
- •Any MCP-compatible tool
Plus our REST API works with any application you build.
Can I delete or edit memories?
Yes! You have full control over your data. Delete individual memories, entire projects, or workspaces at any time. You can also edit memory content to correct mistakes or update information. Your AI will use the latest version.
Next steps
Ready to get started?
Set up ContextStream in 5 minutes and give your AI a persistent memory.