Cowork: JSON Parsing Failure Mid-Response — Fix

A chat reply stops mid-sentence with an error like JSON parse error or failed to parse response. This usually happens when Cowork is converting connector data into a structured artifact (a table, tracker, or spreadsheet). The connector returned data Cowork expected to be valid JSON, but one field broke the parse. The fix is usually a simple retry or narrowing your request scope. Both take under a minute.

Quick Fix

1. Click the retry button below the errored response (or press the up arrow to re-run the same prompt).
2. Cowork re-executes the same tool call against the connector.
3. Wait for the response to complete.

If the retry succeeds, you're done. If the parse error repeats, try the "Narrow the Request Scope" fix below.

Intermittent Parse Error (Most Common)

How ScaleIT handles this: See AI Quick-Start

Connectors occasionally return data in an unexpected format, usually because a single field contains unusual characters, nested structures, or missing values. On retry, the connector returns data in the expected format and the parse succeeds.

Fix: Retry the same prompt. Cowork can retry automatically on most parse failures, but a manual retry (using the button or re-running the prompt) forces a fresh tool call. Most intermittent failures clear on the second pass without any other action.

Request Scope Too Large

If a parse error repeats on the same prompt, the connector is likely returning too much data for Cowork to safely structure in one pass. A year of Slack messages, every file in a Drive folder, or a thousand-row table can exceed the parser's safe limits.

Fix: Narrow the request. Replace vague requests like "get all Slack messages" or "list every file in Drive" with tighter scopes:

• Instead of: "Summarize all messages in #general"
• Try: "Summarize messages in #general from the last 30 days"

• Instead of: "Export all files from Drive"
• Try: "List files created in the last 7 days in the Marketing folder"

Once the structured result lands cleanly, you can widen the window. Knowing the shape of the data on a small pull makes it easier to ask for larger pulls safely.

Malformed Connector Data

Rarely, a connector returns data that is genuinely malformed: a field that's supposed to be a number contains text, a timestamp is in an unexpected format, or required fields are missing. This is usually a bug in the connector itself, not in Cowork.

Fix: Report the failure via the thumbs-down button below the response. The Cowork team uses those signals to harden the parser against the specific tool call that misbehaved. Include the exact prompt and the error message in your report.

Verify

• Retry or narrowed request completes without a parse error banner.
• The structured artifact (table, tracker, spreadsheet) appears and looks correct.
• Column names and data match what you expected to retrieve.

Still Failing?

If parse errors repeat after a retry and narrowing the scope:

1. Copy the exact error message (e.g., JSON parse error at line 4, column 12: unexpected token 'undefined').
2. Note the connector name and the exact prompt you ran.
3. Click the thumbs-down button below the response. Include the error string and prompt in your feedback.
4. Contact ScaleIt if you need to escalate. We can investigate whether the connector is returning malformed data consistently and coordinate a fix with the vendor.

Spending time retrying parse errors instead of building automations? ScaleIt can help optimize your connector data pipelines. Contact us for a free consultation.