Kyanite
Labs
Kyanite Labs is Simon Gonzalez de Cruz's creative development lab for building useful AI tools, MCP servers, media systems, domain software, and learning experiments. Most of the work is open source. The paid path is implementation and advising when someone wants help using a Kyanite tool without doing all the setup themselves.
AI agents should edit video, not just talk about video.
Kyanite builds tools from real creative and technical workflows: MCP servers, Python libraries, CLIs, effect pipelines, demos, docs, and install layers that make the work usable outside the original machine.
The repos are the lab notebook and the product shelf.
Kyanite is build-first. The public repositories show what exists, what is being learned, and which tools are ready for people to inspect, install, adapt, or ask for help implementing.
Git repository archaeology framework for mining commit history, detecting signals, running 6 analysis vectors, and generating engineering reports.
Media MCP MV mcp-videoVideo editing MCP server for AI agents with 87 FFmpeg and Hyperframes tools, Python client, and CLI.
Estimation MCP EP EpochTime estimation MCP server for PERT, COCOMO II, Monte Carlo, sprint forecasting, token-to-time mapping, cost estimation, and schedule risk tools.
Localization MCP DX DialectOSSpanish dialect localization MCP server and CLI across 25 regional variants with register control, structure preservation, and QA gates.
Personal OS IN InnerscapePersonal growth OS in TypeScript for journaling, emotional check-ins, habits, goals, tasks, sleep logs, decluttering, and self-awareness workflows.
Domain Software OG openglazeFree open-source ceramic glaze calculator, UMF analyzer, CTE estimator, recipe manager, and studio tool for potters and ceramic artists.
Learning Diagnostics DA Dev Learning ArchaeologistForensic git-history learning diagnostic for AI-assisted developers that turns commit history into evidence-backed study plans and HTML reports.
The thinking should be as public as the tools.
Build logs, learning notes, agent-system essays, product notes, and postmortems. This is where the videos, repos, and experiments become durable public memory.
mcp-video is a video editing MCP server that gives AI agents direct handles on timelines, effects, FFmpeg, and finished media.
Infinite monkeys, LLMs, and the room around the machineThe argument behind the video: output quality is not just probability. It is architecture, filters, and human taste.
What a working AI tool needs before people can use itA practical checklist for turning a working tool, workflow, or rough app into something other people can understand, install, and use.
Build. Learn. Publish.
The pattern is simple, but not soft: find a real workflow, build a tool around it, write down what happened, and keep improving in public.
Build from real use
Start from a workflow that actually hurts or fascinates: video editing, estimating time, localization, glaze math, learning from commits.
Make it inspectable
Shape the MCP, CLI, app, README, examples, screenshots, tests, metadata, and AI-readable discovery layer so the work can survive outside the chat.
Help people use it
Keep the tools open where possible, sell practical implementation help when someone wants the setup, adaptation, or advising done with them.
Open-source first. Paid help when you want it implemented.
Kyanite Labs is the build lab. PuenteWorks is the separate consulting container. The paid work here is implementation and advising around Kyanite tools, products, workflows, and artifacts.
Public Tools
MCP servers, CLIs, domain tools, experiments, and build notes that you can inspect, install, fork, learn from, or use as starting points.
- mcp-video, Epoch, DialectOS, openglaze, devarch tools
- Repos, docs, examples, release notes, and demos
- Built in public as the Kyanite portfolio and learning trail
Implementation Support
For people who want to use a Kyanite tool without doing all the setup, adaptation, docs, or workflow integration alone.
- Install, configure, and adapt the tool
- Map the workflow and advise on tradeoffs
- Leave docs, examples, and a usable handoff
Agentic Media System
A repeatable short-form video and proof pipeline using content strategy, mcp-video effects, captions, audio handoff, review loops, and publishing assets.
- Built from the Infinite Monkey production workflow
- Visual system, script, caption, and edit handoff
- Turns social clips into durable blog and proof assets
DialectOS Implementation
Help using DialectOS for Spanish docs, app strings, support macros, or locale files across target dialects.
- Up to 10,000 source words and 5 dialects
- Severity table and implementation-ready report
- For teams avoiding embarrassing generic Spanish
Learning Diagnostics
A source-history diagnostic for AI-assisted developers who want evidence about their learning patterns, repeated failures, and next practice loops.
- Uses devarch-framework and learning-archaeology patterns
- Finds repeated failures, habits, and study opportunities
- Outputs a report people can act on
Operator Assets
Downloadable prompts, repo structures, Claude Code workflows, and implementation templates for builders who already ship.
- Claude Code blueprint and productivity pack
- Designed as workflows, not prompt confetti
- Good entry point before scoped work
Bring the messy truth.
Tool you want to use, repo you want installed, workflow you want adapted, docs you do not understand, half-working MCP setup, or a build you want advice on. If it connects to Kyanite's tools and practice, bring the messy truth.
- Best for builders who want implementation help around Kyanite tools.
- Not a fit for generic consulting, empty lead-gen theater, or unrelated agency work.
- PuenteWorks is the separate consulting container; Kyanite is the lab.