Expose per-message timestamps to Claude's context (Cowork + Chat gap — no workaround available)

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Claude Code has no awareness of elapsed time between messages. This has been filed multiple times (#32566, #49084, #18582, #24349 and others). This request adds two missing specifics:

1. The Cowork/Chat gap is unaddressed.

CLI users have a community workaround (claude-context-tick, a UserPromptSubmit hook). Cowork and Claude.ai Chat have no equivalent — hooks are not exposed. For users running structured workflows on Cowork, there is currently no workaround.

1. Concrete downstream failures in a structured personal OS.

I run LCOS (a personal life operating system on Obsidian, driven by Claude in Cowork). Two specific failure modes caused by the timestamp gap:

Changelog insertion disorder. lcos-close appends session entries to a markdown changelog. Without a mid-session timestamp anchor, the insertion point is determined from a stale cached read — if another session ran concurrently, the entry lands out of chronological order.
Energy mode mis-assessment. lcos-energy re-assesses cognitive load at 4–6h elapsed. Without timestamps, elapsed time is estimated, not measured. A session paused for 3 hours looks identical to an uninterrupted flow.

The current workaround is a bash date call at session start (injected by lcos-start). This gives one anchor point. Mid-session gaps — pauses, interruptions, multi-hour context drift — remain opaque for the entire session.

Proposed Solution

Inject ISO 8601 timestamp with each user message turn in Claude's context. The data already exists in the .jsonl transcript — this is a surfacing problem, not a data problem. Reference: #32566 proposed this with full detail.

Alternative Solutions

A lightweight get_time tool Claude can call on demand — zero context cost when unused, available when needed.

Priority

Medium - Would be very helpful

Feature Category

Configuration and settings

Use Case Example

Example 1 — Wasted effort, no self-correction
User: let's build this feature
[45 minutes of back-and-forth, approach not working]
User: we've been going in circles, try something different

With timestamps: model sees 45 minutes elapsed with no progress,
flags the loop and proposes a pivot before the user has to intervene.

Example 2 — Post-compaction context mismatch
Conversation compacted late at night. Summary mentions "it's 1am,
you have an 8am meeting." User returns next afternoon.
Model: "you should get some sleep" — it's 3pm.

With timestamps: model sees 14-hour gap, knows the session-start
context is stale, adjusts accordingly.

Example 3 — Resumed session, wrong assumptions
User works on a task, steps away for 3 hours, returns.
Model continues as if no time passed — same urgency, same context.
User has to re-orient the model manually every time.

With timestamps: model detects the gap, offers re-orientation
prompt ("you were away for ~3h — still on the same task?").

Example 4 — Pacing mismatch
User responds every 30 seconds during active coding.
Then goes quiet for 20 minutes (running tests, reading docs).
Model keeps generating follow-up suggestions into the silence.

With timestamps: model recognizes the shift in cadence,
holds output until the user re-engages.

Additional Context

Issue #32566 (March 2026, labeled area:core) was closed without a public resolution comment. This request is filed to surface the Cowork/Chat-specific gap and the structured-workflow failure modes that weren't in prior filings.