tokconnectAn agency brief earns client trust when it records the decision, the sample, the evidence, the recommendation, and the acceptance test. Send a completed example beside a blank reusable version. A client should be able to trace each proposal to a query or a public record, tell where the evidence stops, and decide whether to fund the next validation.
Scope the brief before using credits
Ask the client to choose one decision: select a first video test, narrow a creator review list, identify a customer question for a demo, or reject weak angles. Name the audience, market, seed phrases, claim restrictions, source requirements, delivery date, and client approver. Set a sample budget before retrieval. For example: five topic candidates, five public videos for each kept candidate, and up to 20 comments on one selected video.
TokConnect charges one credit per MCP tool call; composite tools may collect more than one upstream source inside that call. Follow-ups and pages are separate calls. A budget line belongs in the brief because it lets the client approve the depth of research before the agency expands a promising lead.
| Work item | Client acceptance | Call budget |
|---|---|---|
| Topic scan | Five results with inputs and fields saved | Agree before run |
| Video review set | Links, captions, counts, and reviewer notes | Approve per kept topic |
| Comment follow-up | Fixed count and individual quotations | Approve one selected video |
| Revision | Clarifies existing evidence; new retrieval is a new scope item | State separately |
Estimate a call budget without charging by the row
A small proposed run could use one keyword_research call, two search_videos calls for shortlisted angles, and two video_comments calls for selected posts. That is five MCP tool calls, or five credits under TokConnect’s call-based accounting. Returning twenty records in a call does not make it twenty credits. A sixth call to fetch another page adds one credit.
Keep the credit allowance separate from agency labor. Watching reference videos, checking a product claim, interpreting a sample, and meeting the client require work beyond retrieval. Quote that work in the project scope. A low tool cost does not imply a five-credit research assignment takes five minutes.
Ask the agent to stop after its approved call allowance and return unresolved questions. Reserve an optional follow-up allowance for a question that would change the recommendation. This is a proposed research budget; the coffee example below does not claim that we completed all five of these calls.
Read the completed example
The downloadable coffee brief uses the September 18 evidence. It records the decision: choose one coffee content test a café can support with a real in-store offer. The selected record is coffee shops and solo coffee, with a search-volume estimate of 4,806,180, listed video count 789, and seven returned values from 10,000 to 4,806,180.
The brief recommends a “solo coffee shop ritual” test only after the café chooses a real drink, setting, and supportable claim. That distinction matters. The stored topic field supports a research lead; it does not prove that quiet seating, fast service, or a particular drink is available. The client receives one proposed test and the exact conditions that must be met before production.
Separate evidence, interpretation, and commitment
Use three labels in every brief. Observed contains dates, inputs, public fields, and quotations. Interpretation explains why an observed field may fit the decision, with the sample boundary. Proposed test states the action the client could take and its proof requirement. A recommendation becomes a client commitment only after the named approver accepts it.
For the coffee example, “the returned final trend value is 4,806,180” is observed. “A solo-ritual angle deserves a limited validation” is interpretation. “Film a solo coffee-shop ritual after location confirmation” is proposed. This structure prevents a writer from presenting a planning choice as research fact.
Write the page your client will read first
Open with the decision and the action you need from the client. Keep query IDs and raw JSON in the evidence appendix. A café manager should be able to approve a location check without learning the MCP protocol.
Recommendation: develop one video about visiting the café alone. The saved topic “coffee shops and solo coffee” gives us a relevant research lead. We propose showing an actual drink and a place to enjoy it, with the café’s name and order details visible. Before scripting, please confirm the drink, where filming can happen, and any opening-hours or seating statements. We have not tested this creative or forecast its performance.
Follow that recommendation with one alternative. If the location cannot support the setting, a brewing demonstration may be easier to produce. Mark it as an alternative that needs its own reference review, rather than presenting it as a second researched winner.
| Client question | Answer for this example |
|---|---|
| Why this idea? | The topic names an activity a café can address with a specific setting and offer. |
| What must we supply? | A confirmed drink, location access, and accurate customer information. |
| What does the agency supply? | The evidence appendix, a proposed treatment, and a list of decisions still open. |
| What would stop the idea? | No suitable setting, unavailable offer, or inability to verify the planned statements. |
Run the review cycle
Deliver a first version with source links and a short decision memo. Give the client a defined review route: accept the selected test, reject it, request an evidence clarification, or approve a scoped follow-up. A factual correction updates the record and notes the source. A new market, new seed, bigger video sample, or fresh comment pull changes scope and needs a new call budget.
Set acceptance criteria at kickoff: every recommendation names a supporting record; every public sample states its size and date; open claims have an owner; the client can locate the next validation; and the agency has supplied both a client-ready summary and the underlying evidence link. The brief is complete when those fields are present, even if the client rejects every test.
Create a client research brief for [decision]. Start with scope, audience, seeds, sample size, credit budget, and acceptance criteria. Record each retrieval with date and inputs. Separate observed fields from interpretation and proposed tests. For each proposal, name proof required, owner, next validation, and whether a new retrieval needs client approval.
Resolve a client challenge without defending a weak finding
A client may ask, “Does 4.8 million mean monthly customers?” Answer the measurement question first: it is a returned search estimate, with no confirmed monthly customer unit. Correct any slide that suggests otherwise. Then explain the actual selection reason, which is relevance to the café’s proposed content.
If the client changes the target from local café visitors to wholesale buyers, reopen the brief. Different buyers require different questions and sources. If they want a shorter explanation of the same finding, revise the summary using the existing evidence. This distinction keeps useful editorial revisions separate from new research work.
At delivery, check that the client can open the source links and download the brief, that proposed actions have named owners, and that the chosen version matches the approval email or meeting note. Record the final decision even when it is “do not produce this idea.”
Use the two downloads
Use the completed coffee example to show the expected standard. Use the blank template for the client’s next decision. Add the client’s agreed budget and review owner before work begins. The reusable form holds the structure; the source records make the conclusion inspectable.
Client decision: choose one validated café test
Approved sample: five topic candidates; follow-ups only for kept candidate
Evidence: retrieval date, input, record link, public fields
Recommendation: solo coffee-shop ritual
Proof before production: chosen drink, setting, and approved claim
Revision rule: clarify records at no new retrieval; approve new research separately
Acceptance: client can trace test to evidence and approve or reject it