MCP server bug

Podio MCP bug report

 

Subject: get_tasks unusable — filter parameters rejected by the API

 

Environment: Podio MCP server, accessed via Claude connector. Observed 31 August and 1 September 2026, reproducible.

 

Summary: The get_tasks tool cannot return results under any parameter combination. The API requires a filter, then rejects every filter the tool exposes.

 

Steps to reproduce and actual responses:

 

get_tasks(completed=false, limit=50) →

HTTP 400 — "Query not restrictive enough, must filter on either org, space, app, responsible, reference, created_by or completed_by"

get_tasks(app_id=13681923, completed=false, limit=50) →

HTTP 400 — "Invalid filter app_id"

get_tasks(space_id=3886280, completed=false, limit=50) →

HTTP 400 — "Invalid filter space_id"

get_tasks(org_id=941398, completed=false, limit=50) →

HTTP 400 — "Invalid filter org_id"

 

Likely cause: the error text in step 1 lists the filter names the Podio API expects — org, space, app, responsible, reference, created_by, completed_by — without the _id suffix. The MCP layer appears to forward its own parameter names (app_id, space_id, org_id) straight through to the API, which then rejects them as unknown filters. A rename in the request-building step would fix it. Note also that the tool exposes no parameter for responsible or reference, so there is no working alternative.

 

Secondary issue in the same tool: get_tasks(sort_by="due_date") returns HTTP 400 — Invalid value "due_date": Must be one of ['completed_on', 'created_on', 'rank']. Sorting a task list by due date is the most common case and is supported by the Podio API.

 

Impact: tasks cannot be read at all through this connector. Writing works — create_task, update_task and complete_task all succeed — so the integration can create tasks it can never read back.

 

Two further observations, lower priority:

 

get_item truncates long text fields to roughly 85 characters on about 19% of items, returning "_truncated": true. It is deterministic per item, not load-related: the same item truncates on every call while other items with longer field content return in full (one item returned ~15,000 characters intact). get_items sets _truncated: true on every response regardless.

nextCursor intermittently returns null while more records exist, forcing a fallback to offset for pagination. limit is silently capped at 25 regardless of the value passed.

 

Hope you can fix this. Thank you.


  • Hello @Tim Rood​ - Thank you for posting on the forums.

     

    The Podio MCP server is currently in beta, and our engineering team is actively working to improve and expand its capabilities.

     

    We are gathering feedback from customers who are using the feature during this beta phase and sharing that feedback directly with the relevant teams. At this time, however, Support does not have a timeline for general availability or broader support coverage.