Building Agent Interface
A native workspace for your agent team.
Try Agent Interface ↗ · LinkedIn · Email · Tools · Experiments · Technologies
Your agents work as a team. You stay in the decisions.
Run Claude Code, Codex and OpenCode in one native workspace. See which agent needs you, follow its progress, and discuss plans, documents and results beside the work.
Try Agent Interface → · Install · Source code
Illustration with example data. Open source · macOS · Linux · Windows through WSL
One install sets up the app and its coordination tools. Open Agent Interface and connect an agent tool you already use to get started.
Senior AI and Full Stack Engineer. I build AI systems from the application interface to the infrastructure underneath, including enterprise assistants, retrieval systems and MCP integrations. Below are more tools and experiments you can run and inspect.
Why Backbone · 2:23
why-agent-backbone.mp4
Overview · 1:23
agent-backbone-overview.mp4
Terminal coding agents working as a team across Claude Code, Codex and OpenCode. Communication, delegated work and continuity across projects.
Agent coordination · Python · CLI integrations
Explore the project · Overview on YouTube · Why Backbone on YouTube · YouTube playlist
A 42-second tour of Entune on a Mac:
entune-reel-v3.mp4
Dictation for macOS, Windows and Linux. Choose a local or cloud speech model, speak into any application, and use a personal dictionary whose corrections consider the sentence.
Speech recognition · Contextual corrections · Desktop applications
Explore Entune · Dictionary demo on YouTube · Archived 0.3.0 tour on YouTube · YouTube playlist
Assistant Runtime supplies conversations, streaming, voice and tool requests for applications. Design Studio shows it in use: an assistant edits Markdown and Mermaid diagrams directly in a document editor, with rendering feedback to help repair diagrams.
Pydantic AI · FastAPI · Socket.IO · Next.js · XState
Assistant Runtime · Design Studio
XState MCP gives coding agents access to a running application's state machines. Inspect actors, context and event history, check which transitions are allowed, and send events through explicit write controls.
Full match · 6:35
decision-pitch-full-match.mp4
22 AI players, two managers, one soccer match. Decision models control the players while language models choose lineups and tactics. The match keeps moving while they think.
Decision models · Multi-agent systems · Real-time 3D
Explore Decision Pitch
Watch a race · 28s
race.mp4
How it works · 4min
explainer.mp4
Decision models race through a world that keeps moving while they think. Each model chooses its next move as soon as the previous answer arrives, making both decision quality and response speed visible.
Decision models · Asynchronous inference · Real-time 3D
Explore Decision Arena · Race on YouTube · Explainer on YouTube · YouTube playlist
A voice assistant with a body. Talk to Charlie while separate models handle the conversation and movement, connected through Assistant Runtime.
Voice interaction · Application tools · 3D animation
Explore Avatar Studio
Python · TypeScript · JavaScript · FastAPI · Node.js · GraphQL · Django · Flask
React · Next.js · Redux · Vue · Nuxt · XState · Flutter · React Native
LangChain · LangGraph · MCP · Pydantic AI · Weaviate · PyTorch · TensorFlow · scikit-learn
Docker · Kubernetes · AWS · Azure · Google Cloud · PostgreSQL · Redis · MongoDB





