MCP custom fields cap limitations
Andrew Minion
My Claude is frequently running into errors using the
clickup_get_custom_fields
tool: one of our custom fields is >40KB (lots of really specific epic names) and the server is truncating that field at column 41,769.Would you consider increasing that limit?
Alternatively, could you add an
include_options: true|false
param on the clickup_get_custom_fields
tool so we can exclude if needed? Option lists are ~90% of that payload and the overwhelmingly common need is just the field ID.More details from Claude:
Bug: clickup_get_custom_fields (with list_id or folder_id) and clickup_get_task (with include: ["custom_fields"]) return an SSE frame whose JSON payload is truncated mid-string. The client fails with:
Failed to parse SSE message: Invalid JSON: EOF while parsing a string at line 1 column 41769
Deterministic — the identical call truncates at the identical column on every retry, across sessions and days. Different calls truncate at different offsets (41769 and 53706 observed), so it is not a fixed size cap. There is no partial result and no error object; the tool is simply unusable for these scopes.
Smaller calls on the same connector succeed (space-scoped custom fields ~15KB, 100-task filter_tasks ~36KB, get_workspace_members ~25KB). Other remote MCP connectors in the same client return 290KB+ responses without issue, so this looks connector-side rather than a client or transport limit.
Contributing factor: this space has large dropdowns — one Epic field with 93 options, Area with 32. List/folder scope stacks space-level fields plus list-level ones, which is where it tips over.
Feature request: add a way to fetch field _identity_ without the option payloads — e.g. include_options: false, or a fields name filter on clickup_get_custom_fields. Dropdown options are the overwhelming majority of the bytes and most callers only need the field ID to write a value. This would sidestep the truncation entirely and cut response size by ~90% for the common case.
Secondary: even with that, oversized responses should fail with a structured MCP error rather than a truncated JSON string, so clients can retry or degrade instead of hard-crashing on a parse error.
Log In