Kapacitor/Collaboration

Share the session, then work together.

Collaboration is not another status update. It is a teammate opening the same agent session, the same spec in progress, or the same hosted agent and moving the work forward.

01 · Session links

Send the session, not a wall of context. A Slack link becomes the shared workspace.

alex 10:42

I started the webhook replay spec but I want another read before the agent writes code.

Kapacitor session checkout-api / webhook replay spec Spec in progress · 23 turns · branch spec/webhook-replay Open session Open spec

nina 10:47

Opened it. The retry states need one more acceptance case before we ask the agent to continue.

The second dev does not reconstruct the conversation from Slack fragments. They open the live session, review the spec in progress, and discuss the next change with the first dev and the agent.

Nina

I added the missing acceptance case to the spec. Can you continue from that version?

Agent

I see the spec update: duplicate delivery after a timeout must replay safely and report the original result.

I’ll update the test matrix first, then implement the handler change.

Grounded · from shared session and spec

02 · Shared hosted agent

Two teammates can drive the same agent. One session. Multiple operators. No context split.

Running Claude Code · hosted agent

Repobilling-service

Branchfix/refund-ledger

Shared withAna, Rafael

Controlsprompt, pause, inspect files

One dev starts a hosted agent and shares it. Another teammate joins the same session, asks questions, and helps steer the agent while the transcript stays in one place.

Ana

Before you touch the ledger write, show us the refund path you found.

Rafael

And run the idempotency test after the patch. That is the one that flakes.

Agent

Refunds enter through RefundRequestedHandler, then write to ledger_entries after the gateway callback.

I’ll patch the callback branch and run refund_idempotency_replay before summarizing.

Shared control · one hosted session

03 · Handoff

When someone has to leave, the work keeps its shape. Another dev can recap the started session and continue.

Kapacitor recap

Jordan / search-index backfill

Goal: Backfill missing index rows without blocking writes.
Done: Reproduced the deadlock and ruled out a larger batch size.
Current state: Failing test: BackfillSkipsLockedRows.
Next: Try chunked reads with SKIP LOCKED, then rerun the migration test.

A dev starts a session, then has to leave suddenly. The next teammate reads the recap, opens the transcript when needed, and continues from the actual state of the work.

Priya

Jordan had to step away. Recap the session and continue from the failing test.

Agent

Jordan reproduced the deadlock and ruled out larger batches. The remaining path is chunked reads with SKIP LOCKED.

I’ll open BackfillSkipsLockedRows, make that change, and report back with the migration test result.

Continued · from recap and transcript

How it works

Session links, shared hosted agents, and recaps all point at the same durable session stream. Visibility and sharing decide who can open it; hosted agents let teammates co-drive live work; recap and session recall let the next dev continue after an interruption. Read the mechanics in Visibility and sharing, Hosted agents, Session recap, and Session recall.

Make your team's sessions shareable.

Kapacitor is in private preview. Bring us the places your team already collaborates around agent work, and we'll help wire them into one session stream.

Rather start a conversation? Talk to the team — we’re building with teams that already use coding agents.

Built by the team behind KurrentDB — event streams in production are what we do. Coding agents just produce a new kind.