Run an AI agent inside a workflow. The agent reads a prompt, calls the tools you allow, and returns reply text plus structured fields you can wire into downstream nodes. It supports streaming replies, durable memory threads, and bring-your-own-key providers.
Status: Alpha. Node ID: rollout.ai.
Configuration
Settings live in the node sidebar in four sections: Prompt, Tools, Memory, and Outputs.
Prompt section
| Setting | Description | Default |
|---|---|---|
| System | Standing instructions for the AI (the system message) | — |
| Prompt |
The user turn. Compose wired inputs with {{ }} templates |
— |
| Model | Model from the managed catalog | Gemini 2.5 Flash |
| BYOK Key (optional) | A Connection with your own provider key. Billed to your account, not ours. Required for OpenRouter | — |
| BYOK Model | The model ID to call with your own key (any model your provider offers). Shown when a BYOK key is selected | — |
| Model Options | Provider generation options (temperature, top_p, …) | — |
| Token Budget | Max tokens across all iterations (0 = unlimited) | 0 |
| Parallel Tool Calls | Run independent tool calls in parallel | true |
Tools section
| Setting | Description | Default |
|---|---|---|
| Enable Tool Calling | Allow the agent to call tools | true |
| Agent Tools | The tools the agent may use | — |
| Max Iterations | Max tool calls per request (1–20) | 5 |
Memory section
| Setting | Description | Default |
|---|---|---|
| Thread Key |
Persist the agent’s context under this key and resume from it next execution. Supports {{ }} templates. Blank = in-memory history only |
— |
| Thread Scope | Workflow (shared by every run), Run chain (loop iterations of one run), or Workspace (global key) | Workflow |
| Thread Mode | Read + append, Read only, or Ignore existing (start fresh) | Read + append |
| Replay Window | Resume with only the last N turns (0 = all) | 0 |
| Replay Token Cap | Cap the resumed context’s estimated tokens, dropping the oldest turns first (0 = unlimited) | 0 |
| Compaction | Off, or Summarize — fold older turns into a summary once the thread grows | Off |
| Compaction Threshold (tokens) | Compact once the thread crosses this size | 16000 |
Models
The Model picker is driven by the platform model catalog. Gemini models are always available on the managed key; other managed models depend on your plan. Selecting a BYOK key hides the managed picker — enter any model your provider offers in the BYOK Model field instead. A model outside the catalog (or outside your plan) is rejected before any provider call.
Inputs
| Input | Type | Description |
|---|---|---|
| prompt | string (required) | The user turn |
| system | string (optional) | System instruction — overrides the System field |
| binary | binary (optional) | Image or file bundle for multimodal prompts |
| turns | stream (optional) | Live user turns — each chunk runs one full tool-capable turn with history kept across turns (for example a Speech-to-Text transcript) |
| agent | string (optional) | Act as a specific agent — its identity, memory, and capability profile apply to this node |
| session_id | string (optional) | Pin a specific memory session key — overrides the Thread Key |
| inputs | any |
Wire anything and reference it in the prompt as {{handle}} |
Outputs
| Output | Description |
|---|---|
| output |
Terminal result object: text, tool_calls, structured, usage, finish_reason |
| text |
Convenience port carrying the reply text (same as output.text) |
| deltas | Streaming reply tokens — wire into a stream consumer for live replies |
| events |
Lifecycle events (before_turn, after_turn, tool_call, thinking, meta) — filter by kind |
| callback | Control hook — wire a handler that returns a verdict per lifecycle event |
| errors | Diagnostics, for example one entry per attachment the model could not read |
| binary | Model-generated media (reserved — empty until providers support it) |
Tool calling
Add tools in the Tools section. Any tool-capable node from the catalog can be a tool — the picker groups them by capability: Records, Gmail, Google Drive, Google Sheets, Jira, Web & messaging, Infrastructure & CI, and Workspace. Another AI Agent can be added as a tool too, delegating an open-ended subtask to a sub-agent (nesting is budget-limited per run).
For each tool, every setting of the underlying node is either:
- AI — the model fills the value. You can mark it required and add a description to guide the model.
- Fixed — you set the value; the model never sees or overrides it. Useful to pin a record type or query source while the model chooses the rest.
Tools that need a Connection get a Connection picker. A custom name and description per tool steer how the model understands it.
With tool calling enabled and no tools configured, the agent gets the read-only Query Data tool so it can answer questions about your workspace out of the box.
Structured output
Declare structured fields in the Outputs section under Model produces. Each entry declares a field the model must generate: an ID, a type, an optional description that steers generation, and a required flag. Each entry becomes its own output port you can wire directly; the full parsed object is also available at output.structured.
When structured fields are declared, a reply that does not parse against them fails the node instead of passing bad data downstream.
Streaming replies
The deltas output streams reply tokens as the model generates them, while tools still run between requests. Wire deltas into the Send Conversation Reply node’s text input to stream the reply live into the conversation that triggered the run. Wire events into that node’s thinking input to show the model’s reasoning as a live collapsible section.
Memory threads
Set a Thread Key to give the agent durable memory: the full context — including tool calls and results — persists under the key and is restored on the next execution. Use it for loop iterations within a run or continuity across runs (for example, a per-customer key templated from the trigger). Use Replay Window and Replay Token Cap to bound how much history is restored, and Compaction to summarize older turns instead of dropping them.
Hooks and approvals
Wire the callback output to a handler node to supervise the agent’s loop. The engine calls the handler at lifecycle points (before a turn, on tool calls) and waits for its verdict: continue, halt (stop the loop, keeping the partial result), inject (add context to the next request), mutate the request or result, retry, or suspend the run for a human decision. A suspended run resumes with the decision once a human responds. Suspension applies to batch runs; a run that is actively streaming cannot pause mid-stream.
Stop conditions
Declarative loop halts — stop after N steps, after a specific tool call, on a token budget, a timeout, or a text match — can be set on the node through the workflow API. They are not editable in the sidebar.
Limits
- Max Iterations: 1–20 tool calls per request.
- Token Budget: optional cap across all iterations; when exceeded the node returns partial results with an error.
- Node timeout: 10 minutes.
Tips
- Be specific in prompts and describe the output you want.
- Declare Model produces fields for reliable downstream automation.
-
Wire upstream values into the prompt with
{{ }}templates instead of copying them by hand. - Lower Max Iterations or set a Token Budget while testing to keep runs predictable.
- Fix tool settings you don’t want the model to control by switching them to Fixed.
Troubleshooting
Empty output
- Check that the prompt is not empty.
-
Check the
errorsoutput for rejected attachments.
Model errors
- “Not in the model catalog” or tier errors: pick a model from the catalog, or use a BYOK key.
- A BYOK key with no model set fails before any provider call — fill in BYOK Model.
- A BYOK key that cannot be resolved rejects the call; there is no fallback to the managed key. Check the selected Connection.
Tool calling not working
- Ensure Enable Tool Calling is on.
- Check each tool’s required fields — a Fixed field needs a value from you.
Structured output fails
- The reply must parse against the declared Model produces fields; the node fails rather than emitting unparsed data. Simplify the fields or add descriptions to guide the model.