Skip to main content

AI Automation

AI automation and agentic workflow development

We build automation that removes repetitive work. That can be a custom pipeline, an agentic workflow, or a knowledge base that searches your own documents. It plugs into the tools your team already uses. Common jobs are routine content, reporting, and document search. We work with small teams and agencies.

Book a free discovery call

You may need this if

  • Repetitive engineering and ops work
  • No-code tools that almost fit
  • AI experiments that never reach production
  • Manual processes that do not scale

What the service includes

  • Custom Python automation pipelines
  • Claude Code plugin and slash-command development
  • Knowledge bases that search your documents (Qdrant and Postgres)
  • LLM workflows for content and data
  • Multi-step agentic automations
  • Integration with your existing stack

How the project runs

  1. 01Map. We find the repetitive work worth automating, and the work that should stay with people.
  2. 02Prototype. We build a working pipeline and test it on real inputs. We use Claude Code, Python, and retrieval where each one fits.
  3. 03Ship. We deploy the automation into your stack with logging. People approve the steps that need a human check.
  4. 04Measure. We track the time the workflow takes and tune it.

Recent work

Muster is the back office we use to run Analog Elk. It connects Claude Code to the same records people use. Our own development tools search our documents. They also run spec-driven workflows with multi-agent review and automated deploys.

See selected work →

Common questions

How is this different from Zapier or Make?
Zapier and Make connect apps through triggers and actions. We write custom software when a workflow needs your own data, AI models, or logic those tools do not cover.
What is an agentic workflow?
It is an automation where an AI model plans and runs several steps. It reads context, calls tools, and makes decisions within the workflow.
Do you build custom AI tools?
Yes. That includes Claude Code plugins, knowledge bases, and internal tools built around your data.
Can you integrate with our systems?
Yes. We connect to your CMS, database, APIs, and queues. The automation works inside the tools your team already uses.
Where does our data go?
Our AI tools reach your systems through MCP connections that you set up and control. You decide which tools and data each one can see. You can change or revoke that access at any time.
What does an automation project cost?
A project starts from $4,500 for one automation, built and tested with your team. It begins with an AI Workflow Audit, from $750, which usually takes one to two weeks. The audit fee is credited in full toward the project if you go ahead. A build usually takes 2 to 4 weeks. Every project gets a written quote after the discovery call.

Automate the repetitive work.

Book a free discovery call

We build automation for your process.

  • Websites and appsWe build websites and apps with Next.js and Directus.
  • Design to codeWe design your brand and website, then build it in Next.js.
  • Directus CMSWe set up Directus CMS, self-hosted or on Directus Cloud.
  • Selected workCase studies from client projects and our own products.

From the blog