AI Agents
Pre-built agents write, refresh, fix and pitch. Build your own for any marketing job and run it across hundreds of pages. Every one works from your brand knowledge, Search Console and AI visibility.
Why most AI content misses
Give the same brief to a general AI chatbot and to an Orchly agent. One fills the gaps with guesses. The other checks your data first.
Write our “Harbor vs Lemonade” renters insurance page
Write our “Harbor vs Lemonade” renters insurance page
Search Console 
AI Visibility 
Google top 10 Pre-built agents
The Content, Activation and Site Health agents watch your rankings, AI answers and site, and queue what’s worth doing. Approve an item and the right agent picks it up.
Finds the topics you should own, writes new articles, refreshes pages that are slipping and publishes to your CMS once you approve the draft.
Finds the best-of lists, guides and Reddit threads where buyers look and your brand is missing, scores each by authority and drafts the pitch or reply.
Audits your site for SEO and AI search issues, from broken links to blocked AI crawlers, and works through the fixes in order of impact.
Describe the job in plain words. Orchly writes the skill, builds the agent around it and connects your Search Console and AI search data, then runs it on any page you hand it.
Custom agents
If you can brief a teammate, you can build an agent. Pick a model, give it your data and tools, attach a playbook, and decide when it runs.
Comparison Page Writer 
Claude Sonnet 5Recommended 
Reddit 1,156
PerplexityBotcompetitorText · one per line primary_keywordText Comparison Page Writer Search Console (GSC) 
Analytics (GA4) AI Visibility Comparison Page Writer Use when writing a page that compares us with one competitor.
Article Writer Run #171717 Write a guide on tracking UTM parameters in HubSpot for marketing ops leads. Around 2,000 words, practical, with setup steps.
Claude Sonnet 4.6 Brand Kit Web search Search Console Review mode Comparison Page Writer 
Publish to a CMS when completeWordPress · harborinsurance.com 
SlackPost to #content when a run finishes Collections
Paste your keywords, URLs or competitors, one per line. Add an agent column and press Run. Every row becomes its own task, and the result lands back in the row.
Keywords to cover this quarter. Article Writer drafts each one for review.
An agent reads the other columns in its row and writes its output into the one you pick.
Search volume, difficulty, cost per click and intent, filled in for every keyword.
Try the first 10 rows, retry the ones that failed, or run everything that’s ready.
Look at the same rows as a table, a board, a gallery or a content calendar.
ChatGPT ?
Get a free analysis of your brand visibility in AI tools like ChatGPT, Gemini, Perplexity and more.
You stay in charge
Every agent reads your facts, voice and priorities before it works. Nothing is published and no site fix is applied until someone on your team approves it.
Your products, prices, audience, voice and priorities, learned from your site, files and chats. No agent writes a word without reading it.
Claude Orchly MCP Refresh our three weakest renters pages and tell me when the drafts are ready.
Started 3 Article Refresher runs. The drafts will wait for your approval in Orchly.
Use cases
Each one is a pre-built agent, a custom agent or a collection, set up in minutes.
Each week the Content Agent finds pages sitting at positions 5 to 15 or losing clicks. The Article Refresher adds what they’re missing and queues the update for review.
Paste 40 competitors into a collection. A custom agent writes each “us vs them” page from your pricing, your rankings and the live search results.
Point the Content Translator at your top 50 URLs with a target language. It localizes each page in your voice and keeps its structure and links.
Run the Meta & Schema Optimizer down a column of URLs. It writes titles, descriptions and JSON-LD, and applies them to your site once you approve.
The Activation Agent finds pages that name your competitors and not you. The Outreach Builder drafts a pitch for each one, ready to send from your own inbox.
A custom agent reads AI visibility and Search Console every Monday and posts what moved, and why, to your team’s Slack channel.
The Link Enricher reads your sitemap and adds internal links and outbound citations, so new posts never go live as orphan pages.
A scheduled agent checks which competitors gained mentions in ChatGPT, Perplexity and Gemini, and drafts a response page for the biggest gap.
Each client has its own workspace, brand knowledge and data. The same agent brief produces work in each client’s voice, from that client’s numbers.
They are AI workers that do SEO and AI search jobs from start to finish: research and write an article, refresh a page that is losing traffic, add links, fix metadata and schema, or draft outreach. In Orchly each agent works from your own data (brand knowledge, Search Console, AI visibility) instead of guessing, and hands the work back for review or publishes it to your CMS.
Every workspace gets the Content, Activation and Site Health agents, which find work: pages close to page one, pages losing clicks, sites that cite your competitors, Reddit threads and technical issues. Then there are the agents that do the work: Super Agent, Article Writer, Article Refresher, Content Translator, Link Enricher, Meta & Schema Optimizer, Outreach Builder and the Site Health Agent. You can use them as they are or duplicate one and change it.
Before an agent writes, it reads the sources you allow: your brand profile, writing instructions, uploaded files, your website sitemap, Search Console, Google Analytics, AI visibility and AI traffic. Prices, features and claims come from your files, internal links come from pages that exist, and priorities come from real rankings and AI answers.
Yes. Name it, pick a model, write its instructions in plain language, and choose what it can read and do. Instructions can reference tools with @, skills with @skill and inputs with {{double braces}}. Then run a test, and put it on a schedule once the output is right. Any marketing job you can brief a teammate on can become an agent.
A skill is a playbook written in markdown: the steps, structure and rules for one job, like writing a comparison page. Agents load a skill when the job calls for it. Orchly ships four (AEO Article Writing, Content Refresh, Content Translation and Link Enrichment), and you can write your own. Update a skill once and every agent that uses it improves.
Use a collection. It works like a spreadsheet: paste your keywords, URLs or competitors one per line, add an agent column, map its inputs to your columns, and press Run. Each row runs as its own task and writes its result back into the row. You can run the first few rows as a test, retry failed rows, or schedule the column to work through a set number of rows each time.
Not unless you let them. Every agent runs in Review or Autopilot. In Review it does the work and waits for your approval before anything is published. Site Health fixes are only applied when someone presses Apply, and each change shows the current and new value. Most teams start in Review and move routine jobs to Autopilot once they trust the output.
Claude Sonnet 5 (recommended), Claude Sonnet 4.6, Claude Haiku 4.5, GPT-5.5, GPT-5 mini, GPT-5.4 nano, Gemini 3.5 Flash, Gemini 3.1 Flash Lite, DeepSeek V4 Flash and DeepSeek V4 Pro. Pick a larger model for research and long-form writing and a faster one for short, repetitive jobs.
Finished content can go straight to WordPress, Webflow, Framer, Ghost or Sanity. Refreshes and link updates change the existing page instead of creating a new one. Agents can also post to a Slack channel when a run finishes, and you can start and approve runs from Claude or ChatGPT through the Orchly MCP server.