Three steps.
One shared memory.

Connect your sources, let iri organize them, plug your AI clients in. That's it.

01

Connect your sources

Upload files or sync Notion, Obsidian, Apple Notes, Confluence. iri converts them into searchable markdown automatically.

  • Drag and drop PDFs, DOCX, slides
  • OAuth or API key for hosted tools
  • Background sync, no manual import
02

iri organizes the memory

Your material is chunked, indexed, and turned into a single shared workspace. Sources stay where they are; iri sits on top as the memory layer.

  • Hybrid keyword + semantic search
  • Decisions, owners, and timelines preserved
  • Workspace-level data isolation
03

AI clients connect over MCP

Claude, Cursor, ChatGPT, and any custom agent read the same memory through one standard protocol. No per-tool integration needed.

  • Read with `get_briefing` / `search_notes`
  • Write back with `create_note` / `update_note`
  • Per-agent API keys with scoped access

Memory that
compounds over time.

Every new note, every agent write-back, every correction makes the next answer better.

Notes accumulate

Every upload, sync, and agent write-back lands in the same workspace. Nothing scattered, nothing lost.

AI clients all see the same thing

Whether your team uses Claude, Cursor, or in-house agents, they answer from one shared memory.

Corrections propagate

Edit a fact once. The next briefing reflects the update across every agent and every reader.

Onboarding gets faster

New hires and new agents pick up the workspace immediately. No tribal knowledge bottleneck.

Set up in minutes.
Use it for years.

Free to start. Connect your first source today.