What we tend
Custom agents.
The repetitive stuff that eats your time — handled by AI agents we build for you. Background tools, not chatbots. You own them. No subscription.
What we've built (real examples).
Photo triage agent
Sorts new shop photos by quality, privacy concerns, subject matter. Routes good ones to processing, bad ones to a review bin. Saves 30+ minutes per photo batch.
Review reply drafter
Reads new reviews, drafts replies in your voice (length-matched, tone-matched, specifics-mirrored), queues them for your approval. You thumbs-up via text, they post.
Content QC agent
Every page draft gets audited before publish: voice consistency, banned-words check, factual claims verified against source data. Catches the embarrassing stuff before it ships.
Schema/SEO audit agent
Weekly scan of the site for technical regressions: broken schema, slow pages, broken redirects, missing meta. One-line summary text if issues found.
GBP post drafter
Generates 6 weeks of GBP post drafts ahead of season. You review/approve/edit. We post on schedule.
Inbox triage
Sorts incoming email by category (marketing/transactional/customer/urgent). Surfaces the 5-10 real items. Archives the rest.
How we build them.
Every agent we build follows the same recipe: AI does the draft, a human approves the final. We never auto-fire-and-forget on anything that talks to customers, touches money, or could embarrass you. The AI is a power tool, not the operator.
The code is written in Claude Code or as Claude SDK applications. It runs on your API keys, on your machine or your server. We document everything. You can take the code and walk away if you ever want to.
Most owners don't want to run it themselves — they hire us to tend it (keep prompts current, fix bugs, upgrade models). But the option is yours. You own the agent.
What repetitive task are you doing manually?
If you do the same thing every week and it bugs you — there's probably an agent for it. Text us with the task. We'll tell you if it's worth automating and what it'd cost.
Text us →Automation and agents FAQ
Are these chatbots?
+
No. We don't build customer-facing chatbots — those almost always make the customer experience worse. We build BACKGROUND agents — things that run while you're not looking. Triage your shop photos. Draft review replies for your approval. Audit your website weekly. Generate GBP post drafts. Things that save YOU time, not things that talk to your customers.
Do I have to subscribe to anything?
+
No. The agents we build for you run on your accounts (Claude API, Anthropic, OpenAI — your choice) using your API keys. You pay the API costs directly to the AI provider — usually $20-$100/month depending on volume. We charge for building the agent. After that, it's yours.
Can I run the agents myself?
+
Yes. We hand over working code, documentation, and a runbook. If you want to run it on your machine or your server, you can. Most owners prefer we tend them — keep the prompts current, fix bugs, upgrade when models change — which is a small ongoing cost. But you OWN the agent either way.
What does building an agent cost?
+
Simple ones (photo triage, review reply draft, content QC): $500-$1,500. Multi-step workflows that touch multiple systems: $1,500-$5,000. Full custom suites (a 5-agent setup for an auto shop, for example): $5,000-$15,000. See pricing for full bands.
Are these reliable?
+
More reliable than humans for repetitive tasks, less reliable than humans for judgment calls. We design every agent with a human-in-the-loop gate — AI drafts, human approves. Auto-fire-and-forget is reserved for genuinely safe operations (file moves, data extraction). Anything customer-facing or money-touching gets a human checkpoint.
What's the difference between this and Zapier?
+
Zapier is great for simple if-this-then-that integrations. We use it where it's the right tool. But for actual AI work — making judgment calls, drafting in someone's voice, triaging content, auditing for quality — you need real agents (Claude, GPT-4 class) wired up properly. That's what we build.