TapNow Agent
Learn how TapNow Agent reads context, coordinates Apps, manages canvas content, and supports creative development and production.
TapNow Agent interprets tasks, organizes context, and moves work forward. It can read the text, attachments, and canvas nodes you provide, identify the deliverable and constraints, and select the appropriate App for the task.
Content created or organized by Agent is added to the canvas, where it can be referenced in later conversations. You can set a clear execution boundary at any time: analyze the material, draft a script, build a storyboard, or pause for approval before generation.
How does a task move forward?
A typical task moves through six stages:
- Read the context: review the brief, attachments, and canvas nodes you explicitly provide.
- Confirm the task: identify the deliverable, audience, specifications, deadline, and constraints.
- Plan the work: state which Apps will be used, what will be created, and where approval is required.
- Run the task: generate, edit, or organize content and add the result to the canvas.
- Revise a specific result: update selected nodes while preserving approved details.
- Prepare delivery: check specifications, organize versions, and prepare final files.
When information is missing or contradictory, Agent asks for clarification. Use Ask mode for higher-cost generation, publishing, or external sync actions.
What can Agent help you do?
Organize source material and requirements
Add briefs, meeting notes, reference material, and canvas nodes to the context. Agent can summarize confirmed requirements, flag conflicting information, and list the questions that still need answers.
Develop creative work
During the pre-production stage, use Brainstorm to develop scripts, characters, worlds, campaign concepts, and visual directions. Approved information can be organized on the canvas.
Run production tasks
Once the direction and specifications are clear, Agent can write a script, break it into shots, create image or video tasks, edit existing content, and process repeated assets in batches.
Revise and deliver
Ask Agent to review packaging consistency, copy accuracy, pacing, and delivery specifications. It can also organize versions, prepare handoff notes, and sync content through available Apps.
What do you, Agent, and the canvas each manage?
| Participant | Responsibility |
|---|---|
| You | Set the goal, choose the direction, approve visual and brand details, check usage rights, and review final files |
| Agent | Read source material, organize information, break down tasks, call Apps, run work, and make revisions |
| Canvas | Store briefs, references, approved standards, work in progress, and delivery files |
When giving feedback, name both the requested change and the details to preserve. For example: “Keep the product position and packaging. Replace only the background with a campsite after rain.”
You approve the creative direction, brand accuracy, usage rights, and final release.
Where should you start?
| Current state | Recommended approach |
|---|---|
| A topic or rough idea | Use Brainstorm to develop a direction |
| A clear goal, references, and specifications | Send a direct task instruction |
| One result needs adjustment | Select its node and define the revision scope |
| You need to compare models or settings | Create a generation node and configure it manually |
| The task involves higher spend, sending, or publishing | Use Ask mode |
How do you write a clear task?
A task instruction should contain five types of information:
- Deliverable: what needs to be completed.
- Audience: who the content is for.
- Context: which attachments and canvas nodes to use.
- Constraints: format, duration, language, and any product or character details that must remain consistent.
- Execution boundary: propose a plan, confirm estimated spend, or begin generating.
Copy this instruction and try it with TapNow Agent
Create a 12-second vertical ad for Xiaohongshu using the selected product image. Keep the bottle text and packaging colors unchanged. Use a clean, bright visual style. First show me a three-shot structure and explain the purpose of each shot. Do not generate video yet.
Before execution, check:
- The correct attachments and canvas nodes are referenced.
- The scope is clear, including whether generation or spending will begin.
- The acceptance criteria are complete.
Correct any mismatch in the task instruction before execution.
What do you need to approve?
Approve the following before generation or publishing:
- The creative direction for the audience and brand.
- The source material that can be used.
- The model, generation count, resolution, and estimated spend.
- The accuracy of people, text, trademarks, and product details.
- The final files and publishing specifications.
What if the result drifts?
Identify the affected node and the specific issue:
- Select the node with the problem.
- State the only issue to correct in this pass.
- List the details that must remain unchanged.
- Reference the attachment or node that serves as the source of truth.
- Generate one candidate and review it before continuing.
Copy this instruction and try it with TapNow Agent
The composition and product position are correct. Fix only the bottle cap, using @Front Product Image as the source of truth. Generate one candidate. Do not change the background, lighting, or aspect ratio.
Recommended settings for your first task
- Use Auto in the model selector if you do not have a model requirement.
- Use Ask mode for complex tasks.
- Add the exact reference nodes to the context.
- Handle one major revision per pass.
- Approve a small number of candidates before generating in batches.
How do you confirm that Agent understands the task?
- The deliverable, audience, and specifications are clear.
- The correct attachments and canvas nodes are in use.
- The App, generation count, and stopping point are defined.
- The revision scope and acceptance criteria are clear.
- The next item that requires your approval is identified.
Next: Chat with Agent