Auto-generated GraphQL schema documentation for the Kanvas Ecosystem API, structured for consumption by LLMs and AI coding agents.

This repository contains a complete, flat representation of the Kanvas GraphQL schema. Each type, query, mutation, input, and enum is a standalone Markdown file. There is no nesting, no code to execute — just structured text you can fetch directly.

dev_docs/    Schema from the development branch (latest, may include unreleased features)
prod_docs/   Schema from the latest production release (stable)

Both directories share the same internal structure:

Core indexes (start here):

Auto-generated sub-indexes (only if needed):

Always follow this order to minimize unnecessary fetches:

1. Fetch _LIBRARY_MAP.md to understand the overall context.
2. Fetch the relevant core index depending on what the user needs.
3. Fetch the specific <TypeName>.md file to verify exact field names, argument types, nullability, and enum values.
4. Only fetch auto-generated sub-indexes when you need filter/column/paginator details. Filter type names follow predictable patterns — you can often construct the filename directly.

Never guess field names, argument types, or enum values. Always verify from the docs before generating any GraphQL.

Development (default):

https://raw.githubusercontent.com/bakaphp/graphql-docs-llm/refs/heads/main/dev_docs/

Production:

https://raw.githubusercontent.com/bakaphp/graphql-docs-llm/refs/heads/main/prod_docs/

To get the full definition of the User type from the dev schema:

GET https://raw.githubusercontent.com/bakaphp/graphql-docs-llm/refs/heads/main/dev_docs/User.md

The SKILL.md file in this repository defines a Claude Code skill called kanvas-api-wizard. When installed, it teaches Claude how to navigate this documentation automatically whenever you work on Kanvas GraphQL integrations.

• Claude Code installed and authenticated
• Access to the bakaphp/graphql-docs-llm repository (public)

1. Locate your Claude skills directory

Claude Code loads skills from ~/.claude/skills/. Create the directory if it does not exist:

mkdir -p ~/.claude/skills

2. Copy the skill file

curl -o ~/.claude/skills/kanvas-api-wizard.md \
  https://raw.githubusercontent.com/bakaphp/graphql-docs-llm/refs/heads/main/SKILL.md

Or copy it manually if you have the repository cloned:

cp /path/to/graphql-docs-llm/SKILL.md ~/.claude/skills/kanvas-api-wizard.md

3. Verify

Restart Claude Code and run:

/skills

You should see kanvas-api-wizard listed.

Once installed, Claude will automatically invoke the skill when you ask questions related to the Kanvas API, for example:

• "How do I create a company in Kanvas?"
• "What fields does the User type have?"
• "Write a mutation to update company settings"
• "Which enum values are valid for the notification channel field?"

Claude will fetch the live documentation from this repository, verify all types and arguments, and include documentation references in every answer.

By default the skill reads from dev_docs. If you need to work against the production schema, tell Claude explicitly:

"Use the production schema"

Claude will switch the base URL to prod_docs for that session.

Documentation is regenerated automatically via GitHub Actions:

• Any push to the development branch of kanvas-ecosystem-api triggers a regeneration of dev_docs.
• Any published release on kanvas-ecosystem-api triggers a regeneration of prod_docs.

The generation script (app.ts) runs a GraphQL introspection query against the live API and writes one Markdown file per type. It requires Deno to run locally:

deno run --allow-net --allow-write --allow-read app.ts \
  --url https://your-graphql-endpoint/graphql \
  --output ./dev_docs

.
├── app.ts                          Deno script that generates the docs
├── SKILL.md                        Claude Code skill definition
├── dev_docs/                       Generated docs for the development schema
│   ├── _LIBRARY_MAP.md             Master navigation index
│   ├── _INDEX_QUERIES.md           Core: queries
│   ├── _INDEX_MUTATIONS.md         Core: mutations
│   ├── _INDEX_ENTITIES.md          Core: object types
│   ├── _INDEX_INPUTS.md            Core: input types
│   ├── _INDEX_CONSTANTS.md         Core: enums
│   ├── _INDEX_INPUTS_FILTERS.md    Auto-gen: query filter types
│   ├── _INDEX_CONSTANTS_COLUMNS.md Auto-gen: column enums
│   ├── _INDEX_ENTITIES_PAGINATORS.md Auto-gen: paginators
│   └── <TypeName>.md  (one per type)
├── prod_docs/                      Generated docs for the production schema
│   └── ...
└── .github/workflows/
    └── update-docs.yml             CI that regenerates docs on demand