Real-Time Collaborative Instagram Planner: Google Docs for Your Grid

Why real-time matters for agencies

Picture the most expensive 30 minutes of an agency week: an AM is on a client call hearing edit requests in real time, a designer is back at their desk waiting for direction. Without real-time editing, the AM Slacks "change the second image to the navy version," the designer reads it 12 minutes later, makes the change, the AM checks back, finds the wrong version, asks again, designer fixes again. The change took 25 minutes and three tabs.

With real-time editing, the AM types the change into the post while the client is on the call, the designer sees the cursor land, makes the visual fix while the AM watches, and the post is finished before the call hangs up. Same change, 4 minutes.

Multiply that across 8 clients, 30+ touches a week, and real-time editing becomes one of the highest-leverage operational changes a growing agency can make.

What real-time co-editing actually looks like

• • Live cursors. You see your teammate's cursor with their name floating above it, in their assigned color. You can see exactly which caption field they are editing.
• • Presence avatars. The top of the editor shows who else is currently in the document. Click an avatar to jump to where they are.
• • Conflict-free merging. If two people type into the same caption simultaneously, both edits are preserved without overwriting. The CRDT determines the merged state deterministically.
• • Offline IndexedDB sync. Work offline, the edits queue locally. Reconnect, the CRDT syncs automatically. No "refresh to save" warnings.
• • Per-post editing. The same model applies whether you are editing a post, a caption, the brand bio, or hashtags.

The technical stack: why Yjs

We picked Yjs after testing three CRDT libraries. The reasons:

• • Production-proven. Notion, JetBrains Fleet, Anytype, and many others use Yjs in production at scale.
• • Small payload. The Yjs binary protocol is highly compact — typical edit deltas are under 100 bytes.
• • Provider-agnostic transport. We run it over WebSockets at /ws/. The same Yjs document can also sync over WebRTC if we ever need P2P.
• • IndexedDB persistence out of the box. The y-indexeddb provider gives offline-first by default.
• • Awareness protocol. Cursor positions and user presence are first-class, not bolted on.

When real-time changes the workflow

These are the workflow shifts agencies report after switching to a real-time planner:

• • Live edit during client calls. AM and designer fix posts while client is talking.
• • Pair-content sessions. Strategist and designer build out a week's grid in a 30-minute pair session instead of async over 2 days.
• • Reactive moment-marketing. When a trend breaks, two people can co-author the response in 10 minutes.
• • Faster onboarding. Senior staff can onboard a new hire by joining the same document and pointing at things live.

The pattern: anywhere you used to schedule a Zoom call to co-edit a Google Doc, you can now do the same in the planner without leaving the tool.

Real-time + reviewer chains: the combination

Real-time co-editing pairs especially well with sequential reviewer chains. The pattern:

1. Designer drafts a post.
2. Stage 1 reviewer requests an edit. Designer is online — they jump in real-time, fix it together in 2 minutes.
3. Stage 1 approves. Stage 2 (AM) requests a wording change. Same thing — co-edit live, approve.
4. Stage 3 (client) approves with a single click on the share link.

Without real-time, each rejection round adds hours of latency. With real-time, the chain runs on a sub-hour cycle when reviewers are simultaneously online.

Performance considerations

CRDTs have a reputation for being heavy. Yjs in particular is well-optimized — a typical PlanMyGrid grid document is 50–500KB of CRDT state, which loads in under 200ms on a typical broadband connection. The WebSocket connection adds about 5KB/s of background traffic when actively editing, dropping to near-zero when idle.

For agencies running on flaky connections (think coworking spaces, hotel Wi-Fi), the IndexedDB persistence is the unsung hero. You can lose connectivity mid-edit and keep typing for 20 minutes; when you reconnect, everything syncs in under a second.