Use ShardStitch with Cline.

A stalled Cline request can leave changed files on disk even when the chat never returns. ShardStitch rebuilds that working state before you reopen Cline, then writes a Cline-readable handoff.

After-death behavior: Partial. ShardStitch can rebuild from disk and feed Cline via .clinerules; Recall relies on generic project state.

Integration paths.

Hand-off in

.clinerules.

This is how the next Cline session receives the recovered working state.

MCP

Manual. Configure cline_mcp_settings.json in VS Code globalStorage.

When supported, Cline can call ShardStitch locally instead of relying only on pasted context.

Session recovery

Partial - ShardStitch still rebuilds from verified repo state and local sidecars; deeper session recovery depends on what the target makes available.

Setup path.

Stop the retry loop.

Record the provider error and check whether Cline changed files before the request stalled.

Build the handoff.

Run ShardStitch in that checkout and write .clinerules. Configure cline_mcp_settings.json separately if you need local MCP calls.

Resume one action.

Open a fresh Cline task with the verified diff, the failed request, and one bounded next step. Recheck tests before accepting the result.

Limits to know.

Cline's visible chat may be incomplete after a provider failure. The handoff can establish what reached the repository, but it cannot reconstruct an unsaved response or prove why the provider returned 503. Keep the exact error and extension version separate from verified file state.

Resume this workflow.

If the error is 503 no healthy upstream, start with the Cline 503 guide. For a request that never completes, use the stuck-request guide. Both routes begin by preserving the current diff before another provider or model change.

Best use case.

Starting a cleaner Cline session with current goal, changed files, and next action.

Failure recovery

What fails first.

The visible symptom is often an API request that never finishes, a 0-token context panel, or a provider retry loop such as 503 no healthy upstream. Cline issue reports document these patterns in API Request loading indefinitely (reported in Cline v3.7.1; fixed via PR #2354), 0 tokens and context window, and an unbreakable retry loop. The exact trigger varies by provider, model, repository permissions, and extension version.

What to preserve before retrying.

Current goal and the files that matter next.
Git diff, recent commits, errors, and partial changes.
Decisions, constraints, failed approaches, and what not to retry.
Tests or checks that passed, failed, or remain unknown.

How to continue safely.

Preserve the exact diff and error output before retrying so the next request does not repeat a dead end. Start Cline again if it is the right tool, or hand the packet to another supported AI coding tool without rebuilding the story by hand.

ShardStitch treats the project as evidence: verified disk facts are kept distinct from inferred claims, so the next agent knows what to trust and what to check first.

Common errors and fixes.

Try the smallest reversible fix first, then preserve the state before changing providers or reinstalling the extension.

API Request loading indefinitely

Check provider status, credentials, model selection, and extension logs. If files changed, save the diff before retrying. See Cline issue #2322 (v3.7.1; fixed via PR #2354).

0 tokens or context window

Inspect checkpoint and workspace permissions, then reopen the task with a smaller request. Treat the visible session as untrusted until the working tree is checked. See issue #1782.

503 no healthy upstream loop

Stop repeated retries, switch provider only after recording the error, and continue from the changed files rather than the old transcript. See issue #4205.

Tool repetition loop

If the same observation or tool call repeats without a state change, stop it and restart with a smaller next action.

Browser prompt injection

Treat fetched webpages as untrusted input. Disable or constrain browser actions when the task does not require them.

FAQ.

Can ShardStitch recover a Cline response that never arrived?

No. It can recover saved repository changes and accessible local context, then mark the missing provider response as unknown. Review the packet before giving it to a new Cline task.

Get ShardStitch

Troubleshoot Cline

Open the researched failure guide, preserve the work that reached disk, and continue from a clean verified state.

More Cline recovery guides.

API failure after a model switch, a 0-token context panel, and a stalled terminal command require different checks.