Lunnoa Automate API (Public)

Creates a project in the current workspace. Projects are organisational containers accessible to all workspace users.

Returns all projects in the current workspace, ordered by most recently updated.

Returns a single project.

Returns the feature flags for this deployment, letting a custom UI adapt to the features that are enabled. Flags reflect the current server configuration, such as storage, vector database, email and SSO settings.

Returns the workflows in the current workspace, most recently updated first. By default only the ID and name are returned per item; use the expansion query parameter to include further fields.

Returns a single workflow. By default only the ID and name are returned; use the expansion query parameter to include further fields.

Returns the full catalogue of app definitions, including their actions and triggers. Custom UIs use this catalogue to resolve executionPath step appId/actionId pairs to names and icons. Each action and trigger exposes declarative UI metadata (inputConfig) describing its configuration fields.

Creates a connection for an app using non-interactive credentials such as an API key or basic auth username and password. OAuth2 connection types require the interactive browser flow and cannot be created through this endpoint with machine credentials.

Starts a new execution of the workflow immediately. The workflow must be active and use a manual or scheduled trigger strategy. The optional inputData object supplies values for the custom input fields configured on the workflow "Manually Run" trigger.

Returns the executions in the current workspace, most recently started first. By default only the execution ID is returned per item; use the expansion query parameter to include further fields.

Returns a single execution. The ID and startedAt timestamp are always returned; use the expansion query parameter to include further fields. Request the executionPath expansion to receive the business-readable step list (step labels, per-step status including NEEDS_INPUT, timings, and loop iterations) that custom UIs render progress from.

Creates a variable in the current workspace. Provide projectId to scope the variable to a project.

Returns all variables in the current workspace, ordered by most recently updated.

Returns a single variable.

Updates variable properties. All body fields are optional; only provided fields are changed.

Deletes a variable. This action cannot be undone.

Returns all connections available to the current user: OAuth connections belonging to the workspace and managed connections assigned to the user. Credential values are never returned.

Returns a single connection. Credential values are never returned.

Unauthenticated workflow trigger. The workflow must have a webhook trigger configured. The JSON request body and headers are forwarded to the trigger. When the workflow responds synchronously, the response body contains the output of the webhook trigger; otherwise the body is empty.

Unauthenticated workflow trigger. The workflow must have a webhook trigger configured. The JSON request body and headers are forwarded to the trigger. When the workflow responds synchronously, the response body contains the output of the webhook trigger; otherwise the body is empty.

Unauthenticated workflow trigger. The workflow must have a webhook trigger configured. Query parameters and headers are forwarded to the trigger as the payload. When the workflow responds synchronously, the response body contains the output of the webhook trigger; otherwise the body is empty.

Unauthenticated endpoint implementing the pause-for-user-input pattern: it resumes an execution paused in the NEEDS_INPUT status by submitting values for the node that is waiting. Identify the waiting node via the executionPath expansion on GET /executions/{executionId} (steps with status NEEDS_INPUT). The endpoint acknowledges receipt immediately and the execution resumes asynchronously.

Creates a queue in the given project. Queues are the platform's human-in-the-loop work surface: items are added to a queue, then processed by workflows or reviewed by people.

Lists queues across the workspace, ordered by priority then creation date. Queues are the platform's human-in-the-loop work surface. Configuration fields (maxRetries, retryDelayMinutes, batchSize) are not included in list responses.

Returns a single queue with its configuration.

Updates queue configuration such as name, retry behaviour, batch size, priority, and active state.

Soft-deletes the queue and deactivates it. Items are retained.

Returns item counts for the queue broken down by processing status.

Marks the queue as active so its items can be processed.

Marks the queue as inactive, pausing processing of its items.

Adds an item to the queue for processing or human review. If an externalId is supplied and an item with the same externalId already exists, the existing item is returned with isDuplicate set to true.

Lists items in the queue with pagination, ordered by priority then creation date. Queues are the platform's human-in-the-loop work surface, so this is the primary endpoint for building review interfaces.

Creates multiple queue items in one request. Duplicate handling is controlled by the options field: skip duplicates, update existing items, or return them unchanged.

Applies the same field updates to every listed item in the queue.

Permanently deletes the listed items from the queue. Items that fail to delete are skipped and the response reports the deleted count.

Returns items that are ready for processing: pending items whose scheduled time has passed (or that have no schedule), and retrying items whose retry time has arrived. Intended for workers pulling batches of work.

Returns items in RETRYING status whose next retry time has arrived.

Lists items in the queue that are in the given processing status, with pagination.

Looks up a queue item by the external identifier supplied when it was created.

Searches items in the queue. The search filter matches against externalId, item data, and metadata; other filters behave as in the list endpoint.

Returns aggregate statistics for the queue's items: counts by status and priority band, average processing time, and retry rate.

Returns a single queue item.

Updates item fields such as data, priority, retry configuration, schedule, and metadata.

Permanently deletes the item from the queue.

Returns the chronological list of errors recorded against the item.

Transitions the item to a new processing status. Only valid transitions are accepted; invalid transitions return a 400 error. Moving to FAILED or RETRYING records the supplied error message.

Associates a workflow execution with the item so processing outcomes can be traced.

Manually requeues a FAILED item by transitioning it to RETRYING. Fails with a 400 error if the item is not in FAILED status or has exhausted its retry attempts.

Records an error against the item. If the item is currently PROCESSING it transitions to RETRYING, or to FAILED once retry attempts are exhausted.

Lists every agent in the caller's active workspace, most recently updated first. By default each item carries only `id`, `name`, and `FK_aiConnectionId`; request additional fields with the `expansion` query parameter.

Returns a single agent. By default only `id`, `name`, and `FK_aiConnectionId` are returned; request additional fields with the `expansion` query parameter. Access is gated by the `agents:use` permission, so agents reachable via a share grant can also be fetched.

Lists tasks visible to the caller, most recently updated first. Builders see every task for agents in the active workspace; grant-only callers see only the conversations they own. By default each item carries only `id`, `name`, and `isStreaming`; request additional fields with the `expansion` query parameter.

Returns a single task. By default the response carries `id`, `name`, `isStreaming`, `lastActivityAt`, and `latestRun` (plus `activeStreamId` when a resumable stream may still be running); request additional fields, such as the conversation `messages`, with the `expansion` query parameter.

Updates the task name or description and returns the updated task. By default the response carries `id`, `name`, `isStreaming`, `lastActivityAt`, and `latestRun`; request additional fields with the `expansion` query parameter.

Permanently deletes a task and its messages, and returns the deleted task record.

Creates a task (conversation thread) for an agent. Part of the Agent Runtime API; access is gated by the `agents:use` permission. When `customIdentifier` matches an existing task for the agent, that task is returned instead of creating a new one. The response carries `id`, `name`, `isStreaming`, `lastActivityAt`, and `latestRun` by default; request additional fields with the `expansion` query parameter.

Sends a message to an agent task and waits for the full agent turn to finish before responding. Part of the Agent Runtime API; access is gated by the `agents:use` permission. Creates the task row when `taskId` does not exist yet. Messages use the AI SDK UIMessage format. Consumer (grant-based) callers are subject to a per-grant hourly message cap (429 when exceeded).

Reattaches to an in-flight agent turn after a client disconnect. Part of the Agent Runtime API; access is gated by the `agents:use` permission. Responds with `text/event-stream` (SSE) carrying AI SDK UIMessage chunk parts from the resumable stream, replaying from the beginning of the turn. Returns 204 when there is no active stream to resume.

Sends a message to an agent task and streams the assistant response as `text/event-stream` (SSE) using the AI SDK UIMessage chunk format. Part of the Agent Runtime API; access is gated by the `agents:use` permission. Creates the task row when `taskId` does not exist yet. Streams are resumable: after a disconnect, reattach with `GET /agents/:agentId/tasks/:taskId/stream-message` (which returns 204 when there is nothing to resume). When a stream is already active for the task, the connection attaches as a follower of the in-flight turn. Consumer (grant-based) callers are subject to a per-grant hourly message cap (429 when exceeded).

Requests cancellation of the active agent turn for a task. Part of the Agent Runtime API; access is gated by the `agents:use` permission. Optionally persists the partial assistant message (AI SDK UIMessage format) received so far so it is not lost. Idempotent: returns success even when no stream is active.

Creates a workspace-wide knowledge base with a dedicated vector index. Embedding model, provider and dimensions default to the platform configuration when omitted.

Returns all knowledge bases in the current workspace, ordered by most recently updated.

Returns a single knowledge base.

Updates knowledge base properties. All body fields are optional; only provided fields are changed.

Deletes a knowledge base, its stored files and its vector index. This action cannot be undone.

Chunks the provided text, generates embeddings and stores them in the knowledge base. Multi-chunk uploads are grouped into a single document.

Returns a grouped document view of the knowledge base. Multi-part uploads appear as a single group document; single-part uploads appear as standalone documents.

Returns a document group with all of its parts (vector references), ordered by part number.

Deletes a document group and all of its parts, including stored files and vector index entries. This action cannot be undone.

Creates an entity of the given type. When no state is supplied, the initial state from the entity type state schema is applied automatically.

Returns a paginated list of entities in the workspace, optionally filtered by type, state, owner or a name search.

Partially updates an entity. Supplied attributes are merged into the existing attribute values; fields not supplied are left unchanged.

Soft deletes the entity. It no longer appears in reads or lists but its data is retained and can be restored.

Moves the entity to a new state. When the entity type defines transitions, the change must be a valid transition from the current state, a reason is required where the transition demands one, and any fields required by the target state must be populated. Every change is recorded in the state history.

Returns the current state, the transitions allowed from it, and all states defined by the entity type. When the type has no state schema, all three are empty.

Imports pre-parsed rows into an entity type. Each column mapping routes a source column to a target field; map a column to "name" (or "entityName") to set the entity name, which is required. With updateExisting and matchField set, rows matching an existing entity on that attribute update it instead of creating a new one. Failed rows are reported per row without aborting the import.

Returns entities of a type as flat JSON rows ready to be rendered as CSV or Excel by the caller. Each row always contains id, name, state, createdAt and updatedAt, plus the requested attribute fields. The format field records the intended file format only; the response itself is always JSON.

Returns the column headers and a sample row for importing entities of the given type. The name column is always first and required.

Returns entity types in the workspace, sorted by name. Disabled types are excluded unless includeDisabled is true.