Skip to Content
NEW

Agentic Coding with TVBO

Your AI coding assistant — Claude, Cursor, Copilot, or a local model — doesn't know what TVBO is until you tell it. We ship a bundle of focused skills that teaches your assistant about the three pillars, the model spec format, backend choices, and this platform. Install once, then every session benefits.

Why ship skills with TVBO?

Context, not boilerplate

Large language models are powerful but oblivious. Without context they hallucinate parameter names, pick the wrong backend, or hand-edit generated code. The skills carry the rules they need.

One source, many tools

The same canonical skill renders as a Claude Code skill, a Copilot instruction, a Cursor rule, or a raw system prompt. Pick the target that fits your workflow — no copy-paste between formats.

Always in sync

Skills evolve as TVBO evolves. Re-running tvbo skills install after an upgrade keeps your assistant aligned with the current API, schema, and backend support.

Pick your path

The skills work for every level. Here's where to start.

BEGINNER

I've never used TVBO

Let the AI assistant teach you. Install the skills, then ask "What is TVBO and how do I write my first model?" — you'll get a guided walkthrough with copy-pasteable code.

Best target: claude-code or cursor.

Beginner guide →
INTERMEDIATE

I use TVBO regularly

You know the YAML format and have a backend preference. The skills stop your assistant from making rookie mistakes (label vs name, value on a state variable, wrong extras for a backend).

Best target: claude-code + agents-md for your project repo.

Install & tune →
EXPERT

I contribute to TVBO

Six maintainer skills ship in the TVBO repository itself — covering git policy, surgical coding discipline, LinkML schema regen, codegen templates, the pytest marker matrix, and the design-review workflow.

Best target: repo-local — they're already in .claude/skills/ when you clone.

Maintainer guide →

Install in 30 seconds

1. Install the package

pip install tvbo

2. Install the skills

# Default: Claude Code, user scope
tvbo skills install

# Or Cursor:
tvbo skills install --target cursor

# Or a system prompt for raw API / local model:
tvbo skills install --target prompt > tvbo-system.md

3. Verify

# Show what would be installed
tvbo skills install --dry-run

# List installed skills
ls ~/.claude/skills/tvbo-*

# Remove cleanly when you're done
tvbo skills uninstall

TVBO marks every file it writes with a managed-by: tvbo frontmatter line. Upgrades overwrite our own files safely; files you've customised are left alone unless you pass --force.

What's in the bundle?

Four user skills land on your machine. Each fires on demand when the conversation matches its topic — they cost nothing when you're working on something else.

tvbo-overview
The three pillars (LinkML datamodel, ontology, Python package) plus the Odoo platform. Fires on "what is TVBO?" -style questions.
tvbo-writing-models
How to specify a Dynamics in YAML or Python. Catches the common errors: putting value on a state variable, mixing LaTeX and Python syntax, name mismatches in the equation RHS.
tvbo-running-simulations
The SimulationExperiment(dynamics).run().plot() pattern, which extra to install per backend (tvbo[jax], tvbo[tvb], tvbo[pyrates], tvbo[julia]), Intel-Mac pinning quirks.
tvbo-platform
How to use this platform from Python — retrieve curated knowledge, manage your account, share SimulationExperiments and SimulationStudys (public, or private with an API key), and load shared records back into Python.

Ready to ship your first agent-assisted TVBO model?

Once your assistant knows TVBO, it can scaffold a Dynamics, pick the right backend, run a simulation, plot the result, and even push the experiment to this platform — all in a single conversation.

Read the full guide View on GitHub