Your platform,
ready for the
agentic revolution.
AI agents are about to be your platform’s busiest users. They need an API they can call, a tool server built for them and documentation they can read; a buyer needs a codebase that passes diligence and a launch with proof. Send a written brief and working software lands in your repository, at a flat monthly fee or for a share of the results.

What you get
Four kinds of work. Three are scoped and paid monthly, with a fixed list of deliverables so you know what is arriving before the first invoice. The fourth is paid from what it produces.
Technical diligence
The checks an acquirer's team will run on your codebase and API, built and run for you first.
- Automated review of the repository, every finding with its file and line
- Security review across 17 controls, ranked by severity
- Data-protection posture: what you store, where, and the gaps
- Open-source licence manifest from your lockfiles, copyleft flagged
- Your documentation walked the way a new developer would walk it
- A monthly report with each finding classed as deal risk, price chip, clean-up or cosmetic
Developer platforms
A public API that AI agents and outside developers can find, sign up for and use without a sales call.
- Agent tool server (MCP) so AI assistants can call the API
- API gateway with self-service sandbox keys
- Developer portal with a reference generated from your specification
- SDKs in TypeScript and Python; Go and Java on request
- Starter repository and a twenty-minute first-call tutorial
- Usage metering and a funnel dashboard, so adoption is measured from day one
Launch material
The evidence that a platform is real, documented and in use.
- Integrations for LangChain, LlamaIndex, CrewAI, Vercel AI SDK and n8n
- Evaluation suite and published benchmarks
- Technical SEO: llms.txt, structured data, developer metadata
- Messaging guide, technical articles, launch copy and screencasts
- Developer, enterprise and pricing sections for your marketing site
- A data room kept current every month
Operations as software
For companies that would rather share the upside than pay a fee: the studio builds the software that runs part of your operations, then runs and improves it.
- Manual, repeated work rebuilt as software, with AI handling the routine cases
- Built, operated and improved by the studio, month after month
- Paid from results: the monthly AI usage fees plus a share of the revenue it produces, with no development fee
- Scope, the revenue share and the exit terms agreed in writing before anything is built
How an engagement runs
Send a brief
What you are building, what you need shown, and read access to the repository or the documentation. Written is enough; there is no discovery call to schedule.
Scope comes back in writing
With acceptance criteria you can check yourself. Nothing starts until they are agreed, and the deliverable list does not change without a written change.
Work lands in your repository
Commits as they happen and a written report each month. Every finding carries the file and the line, so your own engineers can act on it without a call.
You review, in an agreed window
Say what is missing, in writing. For work the studio built, fixing it is part of the work, not a change request.
Pay in one of two ways
Scoped work is a flat monthly fee, with no success fee and no automatic renewal, and the project code is yours from the first commit. Operations work is paid from results instead: the monthly AI usage fees plus a share of the revenue the software produces.
Who is behind it
Jonathan Else
Jonathan founded On the Fly Studios in Tbilisi in September 2026, after building developer platforms and a procedural-games toolchain, and runs it as a registered individual entrepreneur. The studio runs an automated pipeline from written inputs to generated, tested software, which is how one person delivers a whole platform and why nothing waits on a meeting.
- Based in
- Tbilisi, Georgia (UTC+4)
- Registered as
- Individual entrepreneur, Georgia, no. 300482831
- Works in
- English, in writing, across time zones
- linkedin.com/in/jonathan-else

Dragons Crack
Every dragon starts in a shell.
You hatch three kilometres under a village that has never heard of you. Eat what you can kill and take what it knew: every animal down here evolved something worth stealing. Climb when the dark runs out of teachers. 2.6 m of basalt and fire by the time you see the sun.
- Status
- In development. A playable slice of the ascent runs on desktop.
- Lineages
- Five, each explained by the route it took up.
- Follow
- The studio on GitHub

- Thicket Hatchling0.55 m
- Crystal Wyrmling0.90 m
- Eastern Lung1.40 m
- Fire Wyvern1.80 m
- Western Drake2.60 m
Every lineage is playable from the shell to the surface. Heights are read from each model’s shipped manifest.
Get in touch
Send what you are building, what you need shown, and where the code lives. A written brief is enough to start; scope and acceptance criteria come back in writing.
