Skip to content
DevAgentRadarDevAgentRadar
Sign in
Proposals/Upgrade Cline SDK to v0.0.72 and handle durable
proposaldeveloperP5Worth a lookCline

Upgrade Cline SDK to v0.0.72 and handle durable queues plus run.failed

Move integrations to Cline SDK v0.0.72 so queued prompts survive aborts, sessions resume with prior context after hub restarts, failed queued turns emit run.failed, and MCP/LiteLLM/network edge cases stop dropping work.

Why this loop

v0.0.72 changes host behavior around interruption and recovery. Queued prompts are preserved across user aborts, drained after self-aborts, and keep in-window queue edits; stop/full-stop is consistent across hosts; session context survives aborts and hub restarts. Queued turns that fail now report run.failed instead of completing silently—hosts that only watch success paths will miss errors. MCP session creation no longer dies on a hung server; unconfigured stdio servers get a 30s initialize budget; remote SSE MCP can prompt OAuth on 401 and supports pre-registered OAuth clients. LiteLLM traffic uses Chat Completions (not Responses API). Mid-stream network drops before any model output retry; Vertex ADC refresh uses the configured fetch. Checkpoint diffs include previously untracked files and work when git is initialized mid-session. Adopt the release, then wire resume, failure signaling, and MCP auth explicitly.

Proposed actions

  1. Bump the Cline SDK dependency to sdk/v0.0.72 (tag sdk/sdk/v0.0.72) and reinstall so lockfiles resolve the new abort/queue and session durability behavior.
  2. In your host, on user abort or hub restart: reconnect to the same session and assert queued prompts plus pre-abort context are restored—do not open a fresh session or drop in-abort queue edits.
  3. Subscribe to run.failed for queued turns; log and surface the failure in UI/CI, and stop treating a finished queue item with no error event as success.
  4. For remote SSE MCP: handle the OAuth authorization prompt on 401 and configure pre-registered OAuth clients where dynamic client registration is unavailable; confirm hung/unconfigured stdio MCP only burns the 30s initialize budget instead of blocking session creation.
  5. Retest LiteLLM proxy calls (Chat Completions path), proxy/custom-transport Vertex ADC refreshes, pre-output mid-stream network retries, and checkpoint diffs after mid-session git init including files that were untracked at snapshot time.

Agent prompt

Paste into your agent or query via MCP (`get_agent_prompt`) — free, no extra AI cost

Paste into Claude Code / CLAUDE.md task

# DevAgentRadar → Claude Code

You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.

## Context
Assistant: Cline
Proposal: Upgrade Cline SDK to v0.0.72 and handle durable queues plus run.failed
Summary: Move integrations to Cline SDK v0.0.72 so queued prompts survive aborts, sessions resume with prior context after hub restarts, failed queued turns emit run.failed, and MCP/LiteLLM/network edge cases stop dropping work.
Primary source: https://github.com/cline/cline/releases/tag/sdk/sdk/v0.0.72

## Why it matters
v0.0.72 changes host behavior around interruption and recovery. Queued prompts are preserved across user aborts, drained after self-aborts, and keep in-window queue edits; stop/full-stop is consistent across hosts; session context survives aborts and hub restarts. Queued turns that fail now report run.failed instead of completing silently—hosts that only watch success paths will miss errors. MCP session creation no longer dies on a hung server; unconfigured stdio servers get a 30s initialize budget; remote SSE MCP can prompt OAuth on 401 and supports pre-registered OAuth clients. LiteLLM traffic uses Chat Completions (not Responses API). Mid-stream network drops before any model output retry; Vertex ADC refresh uses the configured fetch. Checkpoint diffs include previously untracked files and work when git is initialized mid-session. Adopt the release, then wire resume, failure signaling, and MCP auth explicitly.

## Suggested actions
1. Bump the Cline SDK dependency to sdk/v0.0.72 (tag sdk/sdk/v0.0.72) and reinstall so lockfiles resolve the new abort/queue and session durability behavior.
2. In your host, on user abort or hub restart: reconnect to the same session and assert queued prompts plus pre-abort context are restored—do not open a fresh session or drop in-abort queue edits.
3. Subscribe to run.failed for queued turns; log and surface the failure in UI/CI, and stop treating a finished queue item with no error event as success.
4. For remote SSE MCP: handle the OAuth authorization prompt on 401 and configure pre-registered OAuth clients where dynamic client registration is unavailable; confirm hung/unconfigured stdio MCP only burns the 30s initialize budget instead of blocking session creation.
5. Retest LiteLLM proxy calls (Chat Completions path), proxy/custom-transport Vertex ADC refreshes, pre-output mid-stream network retries, and checkpoint diffs after mid-session git init including files that were untracked at snapshot time.


## Your job
1. Restate the change in one sentence.
2. Propose a minimal plan for my repo (or a throwaway pilot).
3. Implement only what I approve; prefer small diffs and tests.
4. Call out risks (permissions, breaking APIs, cost).

Start by confirming you understood the proposal.
mcpmodelsecurityRelease source ↗

Your loop

This browser · no sign-in · not shared as “you”

Your decision stays on this device. A public tally appears after a few votes.

Originating release signal

Clinesdk/sdk/v0.0.72Aug 9, 2026

SDK v0.0.72

Prompts queued during a turn now survive being interrupted: they are preserved across user-initiated aborts, drained after a turn aborts itself, and edits made to the queue inside the abort window are applied rather than lost. · Stopping a session now has consistent full-stop semantics across hosts · Session context stays durable across aborts and hub restarts, so an interrupted session resumes with the state it had rather than a reset one · Queued turns that fail are now reported as `run.failed` instead of completing silently · +8 more changes

Verified excerpt — the source's own words
- Prompts queued during a turn now survive being interrupted: they are preserved across user-initiated aborts, drained after a turn aborts itself, and edits made to the queue inside the abort window are applied rather than lost. Stopping a session now has consistent full-stop semantics across hosts
- Session context stays durable across aborts and hub restarts, so an interrupted session resumes with the state it had rather than a reset one
- Queued turns that fail are now reported as `run.failed` instead of completing silently
- A hung MCP server no longer takes down session creation, and stdio servers that were never configured get a 30-second initialize budget instead of blocking indefinitely
- Remote SSE MCP servers now surface an OAuth authorization prompt on a 401 instead of failing outright, and remote MCP supports pre-registered OAuth clients for setups where dynamic client registration isn't available
- LiteLLM requests route through Chat Completions instead of the Responses API, fixing calls against LiteLLM proxies
- Network interruptions that happen mid-stream but before any model output are retried instead of failing the turn

Excerpt ends here — this release continues at the source ↗.

Primary source ↗