iri docs

Use cases

iri's wedge ICP is people who already work with AI agents day-to-day, and the small teams those people belong to. Every scenario below is something one of our pilot users actually does.

1. Solo Claude Code user, persistent memory across sessions

Problem. You have Claude Code (or Cursor, or Claude Desktop) running on a real codebase. Every new session starts fresh. You re-explain the project, the constraints, the prior decisions. You paste in the same context every time.

With iri. Connect Claude Code to your iri workspace via MCP. Your agent reads the same notes you write, queries past decisions, and writes new findings back. New session, same memory.

What you do. Sign up. Sync from Apple Notes / Obsidian / Notion if you have notes elsewhere, or just start writing in iri. Add the iri MCP server to your .mcp.json. Ask Claude "what did we decide about caching?" and it'll pull from your workspace.

What iri gives you. Persistent context, without the cost of a 200K-token paste-everything prompt every session.

2. Small team, Slack decisions to durable notes

Problem. Decisions get made in #standup and #decisions, then disappear into scrollback. A month later nobody remembers why you went with option A.

With iri. Install the iri Slack app. Anyone runs /iri-summary 24h in #standup and the day's thread becomes a structured note (Decisions / Action items / Context / Open questions) in your iri workspace. Route specific channels to specific spheres so #standup notes land in engineering/standups and #decisions lands in strategy/decisions.

What you do. Install Slack app from your iri workspace's Integrations tab. Pick a default sphere. /invite @iri in any channel you want summarized. Type /iri-summary when the conversation is done.

What iri gives you. Slack stops being a write-only medium. Decisions accumulate in iri, queryable forever.

3. Mobile brain-dump via Telegram

Problem. Ideas come on the train, between meetings, before sleep. By the time you're back at a laptop, half are gone.

With iri. Connect the iri Telegram bot. Type whatever's on your mind in 1:1 with the bot, no app switching. Run /summary at the end of the day and the day's stream-of-thought becomes a single iri note, organized into Decisions / Action items / Ideas / Notes.

What you do. Workspace → Integrations → Telegram → Generate link. Send /start CODE to the bot. Pick a default sphere (e.g. personal/journal). Then text it whenever something comes up.

What iri gives you. A capture surface that meets you where you already are. Daily journaling without the friction of opening a notes app.

4. Multi-agent, single source of truth

Problem. Claude Code writes the patch, Cursor reviews it, an internal bot deploys it. Each tool sees a different slice of context. Fixes happen in chat that the next agent never sees.

With iri. Every agent connects to the same iri workspace via MCP. When one agent finishes, it writes its findings back to iri (create_note, update_note). The next agent picks up from there. Humans see the same workspace.

What you do. Same MCP setup as scenario 1, applied to every agent on the team. Use spheres to scope what each agent sees (e.g. engineering for the dev agent, support for the CS agent).

What iri gives you. Knowledge accumulates in one place even as your tooling churns. The handoff between agents stops dropping context.

5. Founder + 1-3 ICs, replacing scattered tools

Problem. Pitch decks in Google Drive, customer notes in Apple Notes, technical decisions in Linear, daily logs in Slack. Search means remembering where you put it.

With iri. Sync everything in. Apple Notes / Obsidian / Notion / Confluence push into iri continuously. Slack /iri-capture lets you save quick decisions inline. The unified search box (and your agent via MCP) finds it regardless of origin.

What you do. Add sync sources from the workspace dashboard. Use /iri-find in Slack as a quick lookup. Ask Claude or Cursor open-ended questions about your business, the answers come back with citations to specific notes.

What iri gives you. "Where did I put that?" stops being a question. Every tool keeps working; iri is the layer on top.

What we don't do well yet

Setting expectations honestly:

  • Heavy enterprise wikis. Notion / Confluence are still better for large nested wikis with rich page hierarchies. iri is the agent layer above them, not a replacement.
  • Realtime collaboration. No multiplayer cursors. Edit conflicts are last-write-wins.
  • Group chat summary on Telegram or KakaoTalk. Both platforms restrict bot access to group histories. We support 1:1 brain-dumps; group chat is on the roadmap pending API access.
  • Complex permissions. Workspace + sphere membership covers the common cases. SSO + per-document ACLs ship later.

Want to try it?

iri is in pilot. Solo + 5-person teams can sign up free. Larger setups or paid plans, talk to us:

  • Pilot access: hello@iri-ai.com
  • Tell us your scenario; we'll point at the right starting place and help with the first sync source.