about kindgi
Private, affordable AI agents you can fully control.
Kindgi builds a library of focused AI agents backed by small open-source models we host ourselves. The bet: narrow scope and small models, sized to give you privacy, lower cost, and behaviour that stays put.
What we do
A focused library of agents.
Today the library has five live agents: an SEO and AIEO audit, an AI sitemap generator, a website builder, a deal-document analyzer, and a brand-to-theme generator. Each does one job, completely. None tries to be a settings panel for a hundred other jobs.
New agents join the library as the work demands them. Old ones don't drift — once an agent ships, its behaviour is pinned.
How we do it
Small models, narrow scope, your infrastructure.
We use small open-source models we host ourselves. Each is pinned to one job, in a version that doesn't change under you.
Each agent's job is narrow on purpose. A bounded job runs predictably and stays auditable; the model only sees what the job needs.
Because the models run on infrastructure we — or you — control directly, there's no third-party API in the loop. The text you put in doesn't leave the stack.
Why it matters
Private. Affordable. Yours to control.
Private by construction.
Privacy isn't a checkbox in a settings panel — it's the architecture. Your text isn't logged into a third-party API because there is no third-party API.
Affordable at the scale you'll actually use.
A narrow model trained for one job runs at cents per call. That's what makes always-on tools — audits, sitemaps, content monitoring — fit any budget. We pass the cost through; we don't pad it.
Yours to audit and control.
The agent you ran last quarter returns the same answer this quarter. We pin model versions and don't swap them silently. That stability is what makes the output auditable — and what makes a workflow you build today still work a year from now.
The process
Three ways in.
Free to try.
Every agent in the library runs on www.kindgi.com with no signup and no card. Paste a URL, drop a document, run a sample. Output is real — same model, same prompts as the paid tier — only the inputs are throttled.
Sign up to use.
SoonA paid account for the same agents, sized for real work — private and secure. Inputs stay on infrastructure we control and run through open-source small models we host. No third-party LLM sees your data. Pricing will be transparent and per-agent, not enterprise-opaque.
Custom agents.
For teams who need an agent shaped around their own knowhow — your data, your training, your context, your tools. A private CIM analyzer for a deal team. A content agent that knows your brand voice. Internal search across a corpus that can't leave your network. Self-hosted on your infrastructure, or hosted by us with the same boundary. Engagement starts with a call.
Who's behind this
A small team that ships.
Katrin Shechtman
FounderSeventeen years building distributed systems and ML platforms. Most recently Head of Architecture at Thomson Reuters. Earlier: engineering leadership at Lightbend and Paytm Labs. Taught Big Data at the University of Toronto. Speaker at Scala Days and Devoxx.
We grow with the work, not on headcount.
Get in touch