An example of an agent you can build in Sylvie: it writes and rewrites title tags and meta descriptions across a whole site, respecting pixel limits, search intent and your brand's actual voice. It reads each URL, the page content and the keyword you're targeting, then drafts variants a senior SEO would ship. Because it runs on your brand brain, it already knows your product naming, your legal wording and the phrases you have decided never to use.
Sylvie does not ship this agent pre-built. It is an example of an agent your team can build with Sylvie, powered by your brand brain, so it runs your workflow the way you actually run it.
Meta Title & Description is an example of an agent a marketing team can assemble inside Sylvie to handle one of the most repeated, least glamorous jobs in SEO: writing hundreds of title tags and meta descriptions that are unique, on-brand and genuinely click-worthy. Instead of a generic tool producing keyword-stuffed strings, this agent works from your own brand brain, so a product page for a fintech client never sounds like a product page for a beauty client.
Sylvie does not ship it pre-built. You describe the job (rewrite the 400 category pages, follow this keyword map, never say "best" for regulated clients) and Sylvie assembles the agent on top of the guidelines, glossaries and past SEO work already stored for that brand. It stays yours: you edit the rules, and every future run reflects them.
The agent pulls the source of truth first. It reads the brand brain for tone of voice, approved product naming, banned claims, target markets and any SEO decisions your team documented before (for example, "we always lead with the city name on local pages"). Then it reads the page itself: H1, body copy, product attributes, and the keyword or intent you assigned to that URL.
Inside Sylvie it drafts titles and descriptions per URL, checks rendered pixel width instead of naive character counts, flags duplicates across the set and marks pages where the copy contradicts what the brain says about the product. You review in a table, approve or edit inline, and approved wording is written back to the brand brain so the next run and the next teammate inherit the same standard. Export to CSV for your CMS or hand it to the account team.
Every title inherits the tone, naming and claim rules stored in that client's brand brain. A regulated client keeps its careful wording, a challenger brand keeps its edge.
Work that used to eat two days of a junior's week gets drafted in one pass. Your SEO lead reviews and approves instead of typing.
It measures rendered width rather than counting characters, so titles stop getting truncated in the SERP and descriptions keep their call to action visible.
Every draft is compared against the rest of the set and against what is already live, so near-identical category pages stop cannibalising each other.
Agencies build one agent per client, each on its own brand brain, so ten accounts get ten consistent voices instead of one house style flattened over everything.
Edits made in review go back into the brain as decisions. The agent stops repeating the phrasing your team keeps rejecting.
These are the situations where marketing teams and agencies typically build this agent first. Each one is a real, recurring job already sitting on an SEO roadmap.
A retail client replatforms and every category and product page needs fresh metadata before launch. The agent drafts the full set against the new URL map and the client's product naming rules, so the SEO lead spends two afternoons reviewing instead of three weeks writing. The site goes live without a single placeholder title.
You pull pages with high impressions and poor click-through from Search Console and hand the list to the agent. It rewrites titles with a sharper angle drawn from what the brand brain says actually converts for that audience, and you ship 120 rewrites in one sprint instead of picking off ten a month.
A dental group with 74 clinics needs unique metadata per location, not one template with the city swapped. The agent varies the angle per page using the services and opening details in the brain, so the pages stop reading as boilerplate and stop competing with each other in local results.
The agent runs on every draft that reaches the content calendar, so no article ever ships with the CMS auto-generating a description from the first paragraph. Editors get two title options and one description already matched to the target query before the post enters review.
You need a visible win fast. The agent audits existing titles against the newly built brand brain, flags every page with missing, duplicate or truncated metadata, and turns it into a prioritised rewrite sheet you can present in the kickoff deck.
A SaaS client expands into Europe and needs metadata per locale, not translated strings. The agent works from market-specific tone and terminology in the brain, so German pages use the formal register the client insists on and Spanish pages use the product names actually used in that market.
No. It is an example of an agent your team builds inside Sylvie. You define the job, the rules and the review step, and Sylvie assembles it on top of your brand brain. That is exactly why the output sounds like your brand rather than a generic generator.
At minimum: tone of voice, approved product and feature naming, and any claims or words you cannot use. The more the brain holds (past SEO decisions, positioning, audience notes) the sharper the titles get, because the agent stops guessing what matters to that audience.
Yes, and most agencies do. Each client has its own brand brain, so you build the agent once and point it at whichever brain you are working in. Nothing leaks between accounts, and each one keeps its own naming, claims and preferences.
The opposite. Nothing publishes itself: the agent produces a reviewable set and your SEO lead approves, edits or rejects each line. Your edits are stored back in the brain as decisions, so the agent moves closer to your standard on every run.
A chatbot starts from zero every time and has no idea what your client sells, what legal has banned or what you tried last quarter. An agent built in Sylvie starts from the brand brain, works across a full URL set at once, checks pixel width and duplicates, and keeps its decisions where the whole team can see them.
Yes. The agent is yours to edit. If the client decides to lead with price, drop a tagline or add a market, you update the rule or the brain and the next run follows it. There is no vendor roadmap standing between you and the change.
Book a demo and we will map the workflows worth turning into agents, and show how each one runs on your brand brain.
Request a demo