Configuration
The settings you can configure through the app, grouped by where they live. The app is the source of truth — where a default is shown here, it reflects the current platform default. Most settings have sensible defaults; change them only if you have a reason.
Project settings
Section titled “Project settings”Configured in Project settings within each project.
| Setting | Default | Description |
|---|---|---|
| Project name | (required) | Display name in the projects list and page headers. |
| Slug | (from name) | URL identifier for the project. |
| Description | empty | One-line summary shown in the projects list. |
| Icon | default | Icon shown next to the project. |
| Visibility | (per project) | Controls who in the organization can see the project. |
Process & step defaults
Section titled “Process & step defaults”Defaults that apply to process steps. A start event’s trigger is set on the element; the timeout and retry defaults apply per step and can be overridden on any individual step.
| Setting | Default | Description |
|---|---|---|
| Trigger | (required) | What starts the process, set on the start event: Manual / Timer (schedule) / Message (an inbound email or webhook) / Signal / Process (chained). |
| Default step timeout | 5 minutes |
Per-step ceiling before the runtime aborts and retries. Long-running service tasks get a higher ceiling. Individual steps can override. |
| Default step retries | 3 |
Per-step retry count with exponential backoff. Individual steps can override. |
Agent settings
Section titled “Agent settings”Configured per agent on the Agents page.
| Setting | Default | Description |
|---|---|---|
| Model | (required) | The provider and model the agent calls. |
| Temperature | 0.7 |
Randomness. Lower is more deterministic; extraction and classification tasks often want it lower. |
| Max output tokens | 2000 |
Cap on response size. |
| Timeout | 300 s (5 min) |
Per-call ceiling before the runtime aborts and retries. |
| Retry count | 3 |
Retries on transient errors. |
| Request limit | 20 |
Maximum model requests per invocation — guards against runaway tool-calling loops. |
| Fallback | off | Optionally cascade to a second provider and model if the primary fails. |
Connector (integration) settings
Section titled “Connector (integration) settings”Configured per connector at Settings → Integrations (organization-level).
| Setting | Default | Description |
|---|---|---|
| Name | (required) | Organization-unique identifier referenced from process designs and agent tools. |
| Type | (required) | The connector kind — Slack / Email / SMS / Database / CSV / Google Sheets / File storage / HubSpot / API call / Webhook / Document / Message queue / HTTP polling. |
| Scope | Organization | Available organization-wide, or limited to a specific project. |
| Credential | (per type) | Reference to a secret; connectors never store credentials inline. |
The rest of a connector’s configuration — endpoint, auth, options — depends on its type and is entered in the connector’s setup dialog.
MCP server settings
Section titled “MCP server settings”Configured per MCP server on the MCP Servers tab of Settings → Integrations.
| Setting | Default | Description |
|---|---|---|
| Name | (required) | Identifier referenced from agent tool assignments. |
| Transport | (required) | stdio (bundled tools), http, or sse (streaming). |
| Endpoint | (required for http/sse) | The MCP server’s URL. |
| Auth | optional | Auth configuration for remote servers (for example, a bearer token). |
| Credential | (if auth) | Reference to a secret. |
| Tool prefix | none | Optional prefix applied to this server’s tool names. |
| Discovery interval | 5 minutes |
How often the platform refreshes the tool list from the server. |
| Per-tool timeout | 30 s |
Ceiling per individual tool call (maximum 300 s). |
Function settings
Section titled “Function settings”Configured per function (written in Python) on the Functions page.
| Setting | Default | Description |
|---|---|---|
| Name | (required) | Project-unique identifier. |
| Input schema | (required) | Typed inputs the function accepts. |
| Output schema | (required) | Typed return shape. |
| Timeout | 30 s |
Per-invocation sandbox timeout (maximum 60 s). |
| Visibility | private |
private (project-only) or organization (shared across the org). |
Task settings
Section titled “Task settings”Per-task (user task) settings on the BPMN element.
| Setting | Default | Description |
|---|---|---|
| Assignee group | (required) | Group whose members can claim the task. |
| Assignment strategy | offer & claim | How the task is allocated within the group: offer-and-claim, round-robin, random, or least-loaded. |
| Form | none | A bound form whose fields the approver fills in to complete the task. |
| Completion actions | from the form | The buttons the approver sees, defined by the bound form — for example Approve / Reject / Request info. |
Notifications
Section titled “Notifications”Email notifications are configured per user in Account settings (personal email preferences). There is no per-project notification configuration.