For media buyers and strategists who can review the output and put it to use. Everything runs inside Tegra, with nothing to install.
You bring the product page, product facts or account exports the task asks for. You get campaign files, research or creative briefs for review. You decide the final claims, the imports and every account change.
Each task shows its estimated cost before you start. You pay for what you use, and image rendering is only charged when you ask for it.
See what each agent prepares, what it needs from you and what it costs.
Prepare Search, PMax and Demand Gen files, review account exports, or rewrite a Shopping feed. You review the work and import it yourself.
Bring a product URL for campaign builds, account exports for an audit, or a product feed for rewrites.
Explore Google Ads AgentTurn product facts and customer language into Meta angles, production briefs, scripts and feed-copy files.
Bring a product page and real customer reviews for creative work, or dated metrics for a first-pass funnel review.
Explore Meta Ads AgentSee real Meta copy from a run, how it was reviewed and the finished image, then compare the file formats. What your own run can produce depends on the inputs you bring.
Our team ran the agent on the RESIST product page. It wrote five drafts and revised them after its own review. Two passed the final copy check and three were held back as too similar. One of the two became this finished image.
| Angle | Primary text | Headline and description |
|---|---|---|
Composition made concrete | Two sprays on a cool, clean panel. Inside RESIST's water-based hybrid: graphene-infused ceramic, reactive curing and Insta-Bond technology. Ethos names the composition; the application starts with a thin spread. | Graphene + Insta-Bond, spread thin. RESIST's water-based hybrid. |
What follows the buff | The towel leaves the panel; RESIST's curing step follows. Ethos describes reactive curing in air after buffing and names Insta-Bond technology in its water-based graphene-infused ceramic hybrid. | Buff RESIST. Then it cures in air. Ethos's reactive-cure claim. |
The full CSV has all five drafts, each with its text, headline and description. Every row is marked “Draft - not launched.”
The final check passed 2 drafts and failed 3. It judges how distinct each draft is and also checks for banned phrases and AI-sounding writing. The download includes the failed drafts and why they failed.
The image above was rendered later, on request, from the first draft. The render dropped a comma from the headline and changed an apostrophe, which would need fixing before use.
Nothing was launched or tested, and there are no performance results. This file is a one-off for the demonstration, separate from the standard copy file shown below.
Read the copy checks →Source: Runs by our team · prompt version 1.3.0 · The agent’s unedited copy and image, shown as a compressed preview. All five drafts and the review are in the download. Not a launched ad, not a campaign result and not output from the licensed system.
A run by our team turned a public product page into a likely audience, a proposed angle and a production format. It also noted what it could not find proof for.
| Field | Excerpt |
|---|---|
Likely audience | You want a black everyday sneaker for casual walks in warmer weather. |
Proposed angle | The warm-weather black sneaker |
Production format | 4:5, 1080 × 1350 |
What it could not find | no authenticated customer verbatims |
Rejected line: Cooler than your leather sneakers
.
Reason: no comparative temperature or breathability test.
Why it kept this angle: It makes a concrete material fact relevant to a specific wardrobe decision without claiming black shoes are inherently hotter.
Source: Run by our team · prompt version 1.2.0 · Identifiers removed. The personas and angles are the agent’s own proposals; no verified customer quotes were available for this run.
Our team ran the agent on a real product page. It returned six CSV files: one campaign, two ad groups and four responsive Search ads. Download them and check the wording, the settings and what the validator found.
| File | Rows | What is in it |
|---|---|---|
| 1 | One paused campaign, with an example budget and bidding target | |
| 2 | Two intent groups, both paused | |
| 1 | United States presence targeting | |
| 6 | Exact-match keyword suggestions to validate | |
| 7 | A starter list of negative keywords to validate | |
| 4 | Four paused responsive Search ads |
This demonstration used product facts we supplied. The $50/day budget and $25 target CPA are examples, and the keywords and negatives still need validating. Everything is set to Paused. The agent did not access an ad account or launch a campaign.
We asked for two ads per group, one fewer than the recommended three. The other warnings cover short headlines, a truncated description and a reused headline. All seven are in the download.
Read the validation findings →Source: Run by our team · prompt version 1.4.0 · The agent’s unedited output, with private identifiers removed. A demonstration from supplied product facts: not launch-ready, not a client result and not output from the licensed system.
Supply a product page, market, keywords, budget and economics. The builder prepares the campaign files and tells you what its checks found. You review the files and the import preview.
| File | What to check |
|---|---|
search-campaigns.csv | Paused status, budget, bidding target and whether the margins work |
search-ad-groups.csv | Intent grouping and the requested campaign scope |
search-keywords.csv | Match types, where each keyword came from and the CPC you can afford |
search-responsive-search-ads.csv | Claims, landing-page URLs, distinct messages and character limits |
search-negative-keywords.csv | Excluded terms and conflicts with wanted traffic |
If inputs are missing or a check fails, you may need another pass. Everything arrives paused: nothing launches until you import it.
Prepare a Search package →Source: Current hosted Search builder schema · A preview of the file format. A build can include more files than the ones listed.
Product facts and real customer phrases become a persona, an angle to test and a production brief. You can see where each phrase came from, what the agent wrote itself and which claims still need proof.
| CSV fields | What to check |
|---|---|
image_id, slot, format_id, opener_id | Match each row to its image brief, text format and opener. |
primary_text | Five text options per image: story, list, mechanism, one-liner and objection or offer. |
headline_type, headline, description | Five headlines and five descriptions; check product proof and the message. |
image_id,slot,format_id,opener_id,primary_text,headline_type,headline,descriptionThe 5×5×5 set gives you copy options to choose from, not 125 tested ads. Briefs and image prompts are separate from finished images, which are rendered only when you ask.
Prepare a Meta creative foundation →Source: Current hosted feed-copy schema · A preview of the file format. Your run fills in the rows for your product.
Compare the task, its inputs, and the published estimate before opening the workbench. Expand any task for its brief and required fields.
Research uses public sources and your supplied material. Hosted tasks do not connect to an ad account. Review the required inputs and deliverable before starting.
19 of 19 tasks · published task estimates
| Task / inputs | Agent | Price estimate | Estimated duration | Research |
|---|---|---|---|---|
Search campaign buildPrepare a Search package from a product page: keyword groups, requested RSAs, negatives and extensions. The builder checks supported import rules; review claims, economics and the actual import preview.
| Google Ads Agent | $3 estimated text usage | ~4 min | Public-source research |
PMax asset group buildFull Performance Max asset groups from your product page: 15 headlines, 5 long headlines, 5 descriptions, 20-50 search themes, an image brief pack that goes deep on one angle, video briefs, and the launch settings checklist. Ask for the next 10 image briefs in a follow-up to complete the 20-image slate.
| Google Ads Agent | $4 estimated text usage | ~5 min | Public-source research |
Demand Gen kitA Demand Gen kit per persona and angle: 5 headlines and 5 descriptions per ad group, image and video briefs in the ratios Demand Gen accepts, tCPA and budget set from your numbers, and the kill rule.
| Google Ads Agent | $4 estimated text usage | ~4 min | Public-source research |
Prepare a static creative packRender-ready prompts and Google text assets from your page. Standard modes prepare 20 concepts; matrix prepares 22 prompts. Image rendering is a separate paid action from the pack card.
| Google Ads Agent | $4 estimated text usage | ~4 min | Public-source research |
Search-term waste finderChecks your search-term report against your target and a cut-off you set. You get the wasted terms, the ones to keep watching, and negative-keyword files ready to review.
| Google Ads Agent | $3 estimated text usage | ~3 min | Supplied inputs |
Brand traffic routingPrepare brand-traffic routing: a Manual CPC brand campaign, brand-term negatives and a PMax routing checklist. Review actual account settings before applying it.
| Google Ads Agent | $3 estimated text usage | ~3 min | Supplied inputs |
Shopping feed audit + rewriteChecks your Merchant Center feed against the current field rules, flags missing identifiers and weak attributes, and rewrites your weakest titles, short titles, and custom labels into an upload-ready file.
| Google Ads Agent | $4 estimated text usage | ~4 min | Supplied inputs |
Full account auditAudits the parts of the account your exports cover: bidding, PMax settings, Search structure, feed and Demand Gen. Where an export is missing, you get a checklist of what to pull instead of a guess.
| Google Ads Agent | $4 estimated text usage | ~4 min | Supplied inputs |
Keyword expansion packGrows a handful of seed keywords into a researched plan: intent clusters, CPC headroom per keyword against what you can afford, ad-group mapping, and which clusters to fund first.
| Google Ads Agent | $3 estimated text usage | ~4 min | Public-source research |
Competitor ads teardownCompares competitors’ public ads and pages. You get what they say, what they leave out and positioning ideas to test. The differentiation scores are a planning aid, not a verdict.
| Google Ads Agent | $4 estimated text usage | ~4 min | Public-source research |
Creative foundation from URLTurns product facts and real reviews into personas, desires, distinct angles and a static ad brief with copy. Where there are no customer reviews to draw on, it says so.
| Meta Ads Agent | $3 estimated text usage | ~4 min | Public-source research |
Prepare a static creative packRender-ready prompts and a 5x5x5 feed-copy set. Standard modes prepare 20 concepts; matrix prepares 22 prompts across 11 styles. Image rendering is a separate paid action from the pack card.
| Meta Ads Agent | $4 estimated text usage | ~4 min | Public-source research |
Competitor spy (Ad Library)Review visible Ad Library creatives, status and observed longevity. Receive template observations and adaptation hypotheses. Active status and time visible do not establish spend, profit, a winner or why an ad stopped.
| Meta Ads Agent | $4 estimated text usage | ~4 min | Public-source research |
Feed copy pack (5x5x5)Prepare 5 primary texts, 5 headlines and 5 descriptions per image, tagged by format for a future test. Copy options are not observed campaign results.
| Meta Ads Agent | $3 estimated text usage | ~3 min | Public-source research |
Review mining → openersTurns real customer reviews into personas, desires, and openers that keep the customer's own language - plus the words your customers never use.
| Meta Ads Agent | $3 estimated text usage | ~2 min | Reuse research |
Video script (+ roast)Prepare a video script, beat-by-beat structure, voiceover, opener options and a storyboard with review notes. This is production planning, not a rendered video.
| Meta Ads Agent | $4 estimated text usage | ~3 min | Public-source research |
Launch plan + test matrixYour next test, planned the way the production system runs it: how many cells your budget supports, budgets and bid caps as multiples of your target CPA, and the exact kill and scale rules - decided before emotions get involved.
| Meta Ads Agent | $3 estimated text usage | ~2 min | Public-source research |
Funnel diagnosis (7-step walk)Reviews your dated funnel metrics against your target. You get a first read on where the bottleneck is, and what data would tell an ad problem from a funnel, offer or target problem.
| Meta Ads Agent | $3 estimated text usage | ~2 min | Supplied inputs |
3-Rule Filter copy auditPuts your current ad copy on trial: can readers picture it, can it be checked, could a competitor run it unchanged, does your page back it up - and which phrases a real customer would never say.
| Meta Ads Agent | $3 estimated text usage | ~90 sec | Supplied inputs |
Source: Published task estimates · Text estimates only. Final wallet charges follow actual usage; image rendering is separate.
Three steps. Your account stays under your control throughout.
Choose a task and supply your page, reviews, or account exports. The intake form collects what that task needs.
You give · your source materialThe agent reads what you gave it and drafts the work, marking what is sourced, what is assumed and what is missing.
It returns · research + draftRead the output, download the files, and decide what to import into your account or share with your team.
You decide · what to useA diagram of how a run works. You review the output and make every account change yourself.
Confirm the account, product and data match the task you asked for.
Deal with any flagged claims and assumptions before you use the files or recommendations.
You control imports, budgets, publishing, and the final account changes.
Use a guided assessment when you want a structured brief and one consolidated plan.
One Google Ads Agent review of your setup description. Receive provisional priorities and the exports needed to validate them; this is not a full account audit.
Setup assessmentChoose who does the work.
| Option | What you get | What it costs |
|---|---|---|
| Files, research or briefs from public sources and your own inputs. You review them and put them to use. | Pay as you go from a prepaid balance. Task prices are estimates, and image rendering is extra. | |
| Source code, blueprints, documentation and updates for your team to install and run, with installation help and access to the matching agent included. | $19,997 USD once per system. Provider costs, hosting, ad spend and agent usage are separate. | |
| Ongoing Google or Meta media buying, research, creative production, campaign deployment and optimization. | A monthly service fee plus your ad spend. Each channel page lists the current fees. |
Typically live within 48 hours once account access and required assets are ready. Daily and hourly monitoring, with ongoing testing and optimization. You bring the goals, account access, product facts and brand assets. We handle the work; creative review is there whenever you want it.
See done-for-you services →See client work run on the full system. These results come from accounts we manage, not from the hosted agent.
Google campaign builds need a product page, your market, a budget and your margins. Audits need dated account exports and your target. Meta creative work needs product facts and real customer reviews. Each task lists exactly what it needs. Remove passwords and customers’ personal data before you upload files.
Most tasks are estimated at a few dollars. Prices are estimates, not fixed fees: you pay for what the run actually uses, from a prepaid balance. The workbench shows any free trial and your credits. Images are priced per image and only rendered when you ask. Follow-up requests add usage.
No. The agents work from public research and the files you give them. They never log into Google Ads or Meta, publish campaigns or change settings. You review the files, claims and budgets, then import them yourself.
Research is only as complete as its sources, so check the dates, the market and what is marked as missing. The copy and differentiation scores are there to help you review, not to prove an ad will be profitable. A competitor ad that has run for a long time is not proof that it works.
Stay in the same conversation and tell the agent which file or finding is missing before asking again. Your run receipt and balance history show what was charged; a dropped connection does not always mean nothing was. If a charge looks wrong, email us with the run ID.
See our privacy policy for data handling, or have Tegra run the channel.
Google Ads, Meta Ads and Klaviyo Email/SMS. Buying, creative, copy, pages, testing and optimization, handled by the team that built the AI systems.
Typically live within 48 hours. The clock starts when we have account access and the assets we need.
Keeping execution in your team?
Book a paid working session or explore licensed systems.