Docs · Browse sections

Documentation

Persistent context for AI-assisted technical work.

ContextStream gives your AI tools persistent memory, semantic search, and knowledge graphs — without bloating prompts. Start with MCP (recommended) or integrate directly via the API.

What ContextStream is

A memory and context platform for AI tools.

01

Persistent Memory

Decisions, lessons, and context survive across sessions, teams, and tools — automatically scoped.

02

Semantic Search

Search by meaning. Find the right decision or lesson in milliseconds, not by remembering the right keyword.

03

Knowledge Graphs

Decisions, code, and docs linked into a queryable graph. Trace the reasoning behind any change.

04

MCP-Native

First-class integration with Claude, Cursor, VS Code, Codex, and every other MCP-aware editor or agent.

How it works

Two commands, then ask naturally.

terminal · contextstream-mcp
# 1) Run the setup wizard
# macOS / Linux:
curl -fsSL https://contextstream.io/scripts/mcp.sh | bash

# Windows (PowerShell):
irm https://contextstream.io/scripts/mcp.ps1 | iex

# 2) Ask naturally in your AI tool
"session summary"
"remember we're using PostgreSQL"
"what did we decide about auth?"

Prefer manual setup? See MCP Integration for the per-client config format.

Tools at a glance

~11 domain tools, action-dispatched.

search

Hybrid pattern, keyword, and semantic search across memory + code.

memory

Decisions, lessons, docs, todos — captured and queryable.

session

Recall, lessons, plans, and snapshots from past work.

graph

Knowledge graph, file/code dependencies, and impact analysis.

capsule

Portable, scoped context bundles — share or hand off.

skill

Reusable instruction + action workflows that persist across sessions.

Next steps

Pick a path and keep going.