Automatic Session Initialization
Your AI gets context before you even ask
Stop manually explaining your project. ContextStream automatically initializes context when your AI starts a conversation — including recent memory, decisions, and relevant code.
session_init
Initialize conversation with full context — workspace, project, recent memory, and decisions
session(action='user_context')
Retrieve your preferences and coding style so AI adapts to how you work
session(action='capture')
Automatically capture decisions, insights, and preferences from conversations
session(action='capture_lesson')
Capture mistakes and corrections — prevents repeating the same errors
session(action='remember')
Natural language: "Remember I prefer TypeScript strict mode"
session(action='recall')
Natural language: "What were the auth decisions we made?"
Never Lose Important Context Again
Your AI remembers your decisions, preferences, and project context across sessions and tools
Every decision, preference, and insight is stored as searchable memory. Sign up for Pro to experience persistent memory with no expiration.
Memory Events
Store conversations, decisions, insights, bugs, features, and tasks
Decision Tracking
Track architectural decisions with full rationale and history
Lessons Learned
Capture and surface mistakes to prevent repeating errors
Timeline View
See chronological history of all memory events
Updates with History
Update and refine knowledge without losing the original rationale
Find Code by Meaning
Not just keywords — actual understanding
Traditional search finds 'auth'. ContextStream finds 'how is authentication handled' — understanding intent and returning contextually relevant code.
Semantic Search
Vector-based search that understands meaning and intent
Hybrid Search
Combine semantic understanding with keyword precision
Pattern Search
Regex-based search for specific code patterns
Search Suggestions
Get intelligent suggestions based on partial queries
Scoped Search
Filter by workspace, project, or file patterns
Relevance Ranking
Results ranked by semantic similarity and recency
Connect Decisions to Code
A living map of your system
Link decisions, memory, and documentation into a queryable graph so your AI can answer 'why' with context—not guesses. Pro includes Graph-Lite (module-level import graph + 1-hop impact). Elite/Team unlocks the full multi-layer graph and full graph ingest.
Knowledge Nodes
Create structured nodes from decisions, docs, and insights
Relationships
Connect nodes with explicit relationships and provenance
Graph Related
Find related nodes and adjacent context automatically
Path Finding
Trace how concepts connect across your workspace
Decision Graph
See decisions as a navigable graph for fast recall
Contradictions
Detect conflicting information across knowledge and memory
Bring External Work Into Context
PRs, tickets, docs, and messages in one memory layer
Connect GitHub, Slack, and Notion to stream issues, pull requests, conversations, and documentation into your workspace context. Jira coming soon.
GitHub
Sync repos, pull requests, issues, releases, and webhook events into memory
Slack
Connect channels and messages so discussions show up in search and recall
Notion
Sync pages and databases with smart type detection — tasks, wikis, meetings, and more flow into context
Jira (Coming Soon)
Sync projects and issues (including rich descriptions) into your workspace context
Webhooks
Receive real-time updates when issues, PRs, and Notion pages change
Secure OAuth
Workspace-scoped OAuth with encrypted storage and easy revocation
Understand Your Codebase
Dependencies, impact, and architecture analysis
Before you change anything, know what depends on it. Pro includes Graph-Lite for module-level dependencies and 1-hop impact. Elite/Team adds full call/dataflow/type analysis with deeper impact and full graph ingest.
Dependency Analysis
Query what depends on any file, function, or module
Call Path Tracing
Trace execution flows between any two points
Impact Analysis
Understand what would be affected by changes
Circular Dependencies
Detect and visualize circular dependency chains
Unused Code
Find dead code and unreferenced exports
Built for LLM Workflows
Context building designed for AI assistants
Every response is formatted for optimal LLM consumption. Build context with one call, generate development plans, and create actionable tasks.
AI Context Building
Build comprehensive LLM context with code, docs, and memory
Enhanced Context
Deep analysis with memory integration and graph traversal
Plan Generation
Generate development plans from natural language descriptions
Task Generation
Break down work into actionable, trackable tasks
Embeddings API
Generate embeddings for your own vector search needs
Structured Output
JSON responses designed for programmatic consumption
Organize Your Codebase
Workspaces, projects, and indexed files
Organize your code into workspaces and projects. Automatic indexing keeps everything searchable and up-to-date.
Workspaces
Organize projects into logical workspaces with shared memory
Project Overview
Get summary statistics and key entry points
Automatic Indexing
Code is automatically parsed, chunked, and embedded
Index Status
Monitor indexing progress and health
Statistics
File counts, line counts, complexity metrics
Multi-Language Parsing
Support for JS/TS, Python, Rust, and more
Secure by Design
Your code, your control
Industry-standard security with API key management, data isolation, and no vendor lock-in.
API Key Management
Secure API keys with granular permissions
Data Isolation
Strict tenant isolation for all data
JWT Authentication
Standard JWT tokens for user authentication
Configurable
Extensive environment variable configuration
Security Monitoring
Operational visibility and safety controls
Enterprise Controls
SSO, audit logs, and compliance
Team includes SSO and admin controls. Enterprise adds audit logs and enterprise compliance options. Self-hosting is planned for Enterprise.
Admin & RBAC
Control access for teams with role-based permissions (Team)
SSO (SAML)
Single sign-on for your organization (Team)
Audit Logs
Track access and changes (Enterprise)
Self-Hosting (Planned)
Enterprise self-hosting option for data sovereignty (Enterprise)
~11 Domain Tools, All Capabilities
v0.4.x consolidates all ContextStream capabilities into a token-efficient toolset — no registry bloat, more room for your actual code.
See full action reference for complete documentation.
session_init
Initialize conversation with full context — workspace, project, recent memory, decisions, and relevant code.
context_smart
Semantic search on each message to find relevant context from any time — not just recent items.
search
Search with modes: semantic (meaning-based), hybrid (semantic + keyword), keyword (exact match), pattern (regex).
session
Session management including capture, recall, remember, lessons, summary, and user context.
memory
CRUD operations for events and knowledge nodes, plus search, timeline, and decisions.
graph
Code graph analysis: dependencies, impact analysis, call paths, circular dependencies, unused code.
project
Project management: list, create, index, statistics, and local folder ingestion.
workspace
Workspace management: list, associate folders, bootstrap new workspaces.
integration
GitHub, Slack, and Notion integration: search, stats, activity, knowledge extraction.
reminder
Reminder management: create, list, snooze, complete, and dismiss.
help
Utility operations: list tools, check auth, version info, generate editor rules.
Why consolidated tools?
Large tool registries consume significant context tokens. With v0.4.x, ContextStream uses action/mode dispatch within ~11 domain tools — giving you the same capabilities with ~75% fewer tool tokens.
Not just another memory plugin
Built-in memory resets when you close the chat. ContextStream remembers everything — across sessions, across tools, across your entire codebase. Yesterday's decisions show up in today's context. Switch from Cursor to Claude Code and your AI already knows what you're building.