Shared project memory

Your coding agent forgets every session.

Give it persistent project memory, semantic code search, and the decisions behind your codebase. ContextStream helps the next AI session start with what the team already learned.

Start building freeCompare plans

Install or upgrade in one command

macOS / Linuxcurl -fsSL https://contextstream.io/scripts/mcp.sh | bash
Windows PowerShellirm https://contextstream.io/scripts/mcp.ps1 | iex

macOS, Linux, and Windows · interactive setup · Starter includes 10,000 monthly credits · no card

  • Claude Code
  • Cursor
  • Codex CLI
  • Cline
  • Windsurf
  • VS Code + Copilot
Claude Code~/acme/webContextStream connected
What did we decide about auth retries, and which code depends on it?
contextstream · decisionsquery: "auth retries"
DecisionRetry refresh once, then sign outFeb 10 · active
contextstream · lessonsquery: "auth retries"
LessonA second retry caused sign-in loopswarns on retry changes
contextstream · searchmode: hybrid
Codesrc/auth/refresh-coordinator.ts:112-148indexed 4m ago · current

Claude CodeThe policy is one refresh retry, then sign out. It was decided on Feb 10 because a second retry caused sign-in loops, and it lives in refresh-coordinator.ts, lines 112–148. I'll keep that behaviour and read those lines before editing.

First successful use: ask the question your last session already answered. ContextStream retrieves the relevant decision, code, and project scope so the agent can act from evidence instead of guessing.

01What changes after connection

Thirty seconds from “what is this?” to “why it matters.”

ContextStream turns the context your team normally loses between sessions into shared memory future agents can use before they touch the repo.

01

Capture what the team learns

Decisions, corrections, lessons, runbooks, tasks, and repo guardrails become durable project context.

02

Retrieve it before the next edit

Your agent searches the codebase and the project memory together instead of starting from a blank session.

03

Carry context across tools

The same project memory follows the team between MCP-compatible coding agents and editors.

02Measured proof

23 / 24

agent tasks completed with project memory

Compared with 14 / 24 without it, using the same agent and the same repository. This is a 24-task evaluation, not a customer outcome claim.

Agent tasks completed · same agent, same repo

With project memory23 / 24
Without it14 / 24
A small sample, reported as one. Methods and results →

Stop reteaching agents the same repo context.

Install ContextStream, connect one project, and make the next AI session better than the last one.

Start building freeCompare plans