KoBoToolbox

Connect AI agents to KoBoToolbox, a field data collection and survey platform. Agents manage forms (assets), read and filter collected submissions, delete records, and generate CSV or XLS exports via an account API token.

13 actions available

A scheduled poll or upstream event checks for new field submissions

Agent runs Get Submissions with a Mongo-style query for new records

Agent reads each submission's answered fields and system fields

Agent applies the query filter to separate routine records from cases needing attention

Agent appends routine submissions to the reporting spreadsheet

Coordinator notified in Slack with the count of new and flagged records

Flagging a case for follow-up routed to the coordinator for the call

What This Integration Enables

Agents work across the KoBoToolbox data surface. On the asset side they list forms and projects, read a single asset and its full survey content, create a new asset, deploy it for the first time, redeploy an edited form, and read a deployment's status and submission count. On the submission side they retrieve collected records with an optional Mongo-style query, sort, and pagination, read a single submission by id, count submissions efficiently, and delete a record (permanently, and never on the agent's own authority). On the export side they create a CSV or XLS export task and list export tasks to poll status and retrieve the download link once complete. Because exports generate asynchronously, an agent creates the export, then lists exports until the status is complete before sending the link on. The connector has no triggers, so flows run on a schedule or on an upstream event, then filter and route what the field collected.

Without FlowRunner

Submissions exported on a schedule Someone remembers to pull an export weekly, so data lags behind the field

Records filtered by hand Analysts open the dataset and scroll to find the records that matter

Urgent cases surface late A report that needs immediate attention waits in the dataset until the next review

With FlowRunner

Submissions pulled as they arrive Get Submissions runs on a poll and reads new records without a manual export

Records filtered with a query A Mongo-style query separates routine records from cases needing attention server-side

Urgent cases flagged to a person Records matching the attention filter are routed to a coordinator for follow-up

Use Case Scenarios

Filtered Submission Sync

A survey team collects records in the field throughout the day. On a schedule, the agent runs Get Submissions with a Mongo-style query for records newer than the last run, reads the answered fields, and appends each routine record to a reporting spreadsheet. It reads Get Submission Count to confirm the total matches, then posts a summary to the operations channel: how many new records arrived and how many matched the attention filter. Data reaches the report as the field collects it, instead of lagging a manual export by a week.

Export and Distribution

A program manager needs the week's dataset as a spreadsheet. The agent runs Create Export in CSV or XLS format, then polls List Exports until the export's status is complete and its download URL is ready. Because exports are asynchronous, the agent waits on the status rather than assuming the file is done. Once the link is available, the agent emails it to the stakeholder list and archives a copy. The manual "generate, wait, download, send" loop runs as one flow.

Form Deployment and Attention Routing

A coordinator finalizes a new intake survey. The agent creates the form with Create Asset, activates it with Deploy Asset, and confirms the active deployment with Get Deployment. As responses arrive, the agent runs Get Submissions with a query for out-of-range or high-priority answers, separating the records that need a human from the routine ones. Routine records feed the report; flagged records get routed to the coordinator. Form setup and triage run off the same connector.

Human-in-Loop Highlight

Acting on a submitted field record is exactly where an agent should stop and ask. A flagged case (an urgent report, an answer outside the expected range, a record that fails a data check) is a judgment call, not a routing rule. When Get Submissions returns a record that matches the attention filter, the agent does not act on it alone. It reads the full submission with Get Submission, attaches the answered fields and the reason it was flagged, and routes to the coordinator via Slack: "Submission [id] flagged: [reason]. Full record attached. Open a follow-up, mark it reviewed, or hold for the team?" The coordinator decides. Deleting a record is treated the same way, because Delete Submission is permanent and cannot be undone. Routine sync, counting, and export flow through automatically. Acting on a case, and destroying a record, get a human in the loop.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

13 actions

Assets

List Assets Lists assets (forms and projects) on your account, with the total count and each asset's uid, name, type, owner, deployment status, and submission count. Supports search and pagination.
Get Asset Retrieves a single asset by uid, including name, type, owner, permissions, deployment status, and submission count. Does not include full survey content.
Get Asset Content Retrieves the full definition of an asset, including its survey content (questions, choices, settings). Useful for inspecting or duplicating a form's schema.
Create Asset Creates a new asset (typically a survey) on your account. Returns a draft asset with its new uid. Deploy it with Deploy Asset before collecting data.
Deploy Asset Deploys an asset for the first time, creating an active deployment that can receive submissions. Use Redeploy Asset to push later changes to an already-deployed form.
Redeploy Asset Redeploys an already-deployed asset, pushing its latest saved version to the active deployment. Use this after editing a form that was previously deployed.
Get Deployment Retrieves the deployment details for an asset, including whether it is active, the deployed version, the backend, and the current submission count.

Submissions

Get Submissions Retrieves collected records for a deployed asset. Supports a Mongo-style JSON query filter, sorting, and pagination. Used to pull new records and separate routine ones from cases needing attention.
Get Submission Retrieves a single submission by its numeric id, including system fields and all answered question fields. Used to read a flagged record in full before routing it.
Get Submission Count Returns the total number of submissions for a deployed asset efficiently, reading the count from a minimal page rather than fetching all records.
Delete Submission Permanently deletes a single submission by its numeric id. This cannot be undone, so it is routed for human sign-off before it runs.

Exports

Create Export Creates a data export task for a deployed asset in CSV or XLS format, with optional field and label-language control. Exports generate asynchronously; the response includes the export uid.
List Exports Lists the export tasks for a deployed asset, including each export's uid, format, status, and download URL once complete. Poll this to retrieve the link for an export created with Create Export.

Start building with KoBoToolbox

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us