Working with Tap AI
Use the Tap Agent to create and manage Campaigns, review work, compare responses, prepare private buys, use files and memory, and approve consequential actions.
5 min readUpdated
Something missing from this page? Email help@tap.co
The Tap Agent is the buyer workspace for planning and transaction preparation. Start a new chat from Home, choose a Brand when the work is brand-specific, and describe the outcome in plain language.
A specific campaign brief that asks for media options or pricing normally gives the Agent enough direction to create an editable Campaign and private RFP drafts. You do not need to use command words. Say research only, compare only, or make no changes when you want an answer without workspace edits.
Creating or editing private drafts does not contact a supplier. The Agent should state consequential assumptions, leave unsupported commercial facts open, and distinguish editable planning CPMs and delivery goals from supplier prices or confirmed availability.
A conversation without a Campaign stays chat-only. After the conversation creates or opens a Campaign artifact, wide screens show chat beside it. On a compact screen, switch between Chat and Campaign. The visible Campaign remains the source of truth while the Agent works.
Campaign team chat is separate. It is shared with current Campaign assignees and is not input to the Tap Agent.
For a new Campaign, the Agent can save the grounded objective, budget, flight, markets, audiences, and channels, then prepare the complete set of private broadcast RFP drafts. When the brief supports it, the Agent can recommend editable media, dayparts, spot lengths, budget allocations, planning CPMs, and a response deadline. Review those assumptions before any request is sent.
For existing work, ask the Agent to:
A Campaign header edit by itself does not rewrite its RFPs. When you ask the Agent to carry a supported change through, it updates the Campaign and affected unsent drafts together or leaves them unchanged if the operation cannot complete safely. A Campaign budget change does not reduce an existing Order.
Sent RFPs and Orders stay immutable. A material sent-request change needs a new reviewed revision and explicit resend. A live deadline-only change uses its separate amendment action.
The Agent can compare current submitted response options, projected spend, targeted impressions, CPM, and request fit. It can calculate and save a candidate schedule with whole-spot weekly quantities, then explain gaps and alternatives.
The saved candidate remains private. It does not contact suppliers or create an Order. Order placement is a separate approval action for an authorized buyer. Client proposal sharing is also separate and never places an Order automatically.
Use the model selector in the composer. The current choices are GPT-5.6 Sol, GPT-5.6 Terra, Claude Opus 5, and Claude Sonnet 5. New conversations start with GPT-5.6 Sol.
A different choice belongs to that conversation and is saved with the next message. Changing the model does not change workspace access, available commands, source-grounding requirements, or approval boundaries.
Recent Agent conversations appear in the navigation. Open the chat library to search titles, Campaigns, Brands, folders, and message content; create nested folders; archive or restore a conversation; and organize standalone work.
A conversation linked to a Campaign follows that Campaign folder. Deleting a folder moves its Campaigns, conversations, and subfolders up one level. It does not delete or archive their content.
Select Add files and integrations, then Add photos and files. Choose Team, Me, or the active Brand as the reuse scope. Wait for upload and parsing to finish before expecting the Agent to use the file. The original file and searchable record also appear in Memories.
The Agent should cite the source it used and separate source facts from inference. An upload is evidence, not authority to bypass a permission or approval.
The Integrations sheet currently supports U.S. Census data and one remote Custom MCP server. Custom server tools require approval for every call. A service marked Coming soon is not connected or available to the Agent.
Ask the Agent to save a reusable fact as User, Team, or Brand memory. Open Memories to inspect, search, edit, link, unlink, or forget saved context. Replace an outdated fact instead of keeping conflicting duplicates.
Draft work can proceed inside the buyer workspace. The Agent pauses before consequential actions, including:
Read the action, consequence, and acting organization before approving. Declining the approval leaves the external or destructive action undone.
Use the retry action when it is offered. The conversation and last committed Campaign remain available. If a Campaign update is still running, wait for it to settle before editing the same fields. Retry a failed file from Memories instead of assuming the Agent read it.
Start a new chat for unrelated work. Continue the current chat when you want the Agent to retain its Campaign, messages, and artifact context.