05 / 06MCP · universal integration
Bring ContextStream into the AI tools you already use.
The ContextStream MCP server gives compatible AI tools access to persistent project memory, semantic search, Knowledge Stream, Knowledge Atlas, ContextCapsules, and scoped project context.
# macOS / Linux — install or upgrade in one command
$ curl -fsSL https://contextstream.io/scripts/mcp.sh | bash
# Windows · PowerShell
$ irm https://contextstream.io/scripts/mcp.ps1 | iex
✓ ContextStream MCP installed · scope=streambox/api
01 Works with your stack
One server. Every MCP-compatible tool.
◈
Claude Code
first-class · stable
◈
Cursor
stable
◈
Codex
stable
◈
Cline
stable
◈
Windsurf
stable
◈
Other MCP tools
compatible
02 What MCP unlocks
Seven capabilities. Same scope rules as the rest of ContextStream.
○
Session recall
Pick up where you left off. Sessions span tools.
◈
Semantic search
Query the Stream and Atlas in natural language.
◇
Memory capture
Record decisions and lessons mid-conversation.
◆
Decisions & lessons
Source-backed, scoped, citable.
⊢
Graph traversal
Right context, not most recent.
★
Capsules
Open, fetch chunks, cite back.
03 Natural-language examples
Type as you would think.
› What did we decide about auth?› Remember this lesson: never push without running tests.› Search the project for rate-limit decisions.› Create a handoff capsule for Jordan.› What are the open guardrails on auth/session.ts?› Open cap_8f3c and summarise.
04 MCP vs ContextStream Code
Two surfaces. Pick one or use both.
Where it runs
Inside your existing AI tool
Native ContextStream CLI / panel
Setup
One install command
Built in
Best for
Existing tool workflows
Native first-party experience
Recall surface
Tool prompts
StreamPilot inline panel
Capsules
Open / fetch / cite
Author + open + verify
Multi-tool
Yes — universal
No — first-party only
MCP Server
ContextStream Code
Install MCP.
One command. Every MCP-compatible tool. Your project context, ready in five minutes.