LLM-Assisted Development
An agent-native CMS
Chinaski ships an MCP server, so an AI assistant can manage your whole site — create pages, build blocks, run translations, trigger deploys — through a clean, typed tool interface.
View on GitHubWhy agent-native matters
Your whole site is addressable
An MCP client can list pages, read block content, create and update anything in the CMS, trigger a build, and check build status — all through a typed tool interface without touching the web UI.
Structured content makes good context
Content is stored as structured blocks rather than raw HTML. An AI assistant can read and edit it at the level of intent — update the headline, change the CTA button, add a column — without parsing markup.
Translation is a first-class workflow
The MCP server includes tools for drafting translations via LLM, queuing them for human review, and publishing approved versions. Building a multilingual site is an agent task, not a manual job.
What you can do with an agent
Create and update content
An agent can create pages, add blocks, update headings and body text, set metadata, and publish — all through MCP tool calls. A full site build-out that would take hours in the web UI can be done in minutes with a well-prompted agent.
Run translations
Ask an agent to translate all unpublished Spanish content, review the queue, and publish approved items. The agent calls translation tools in sequence, checks status, and reports what was published.
Audit and fix content
The MCP server includes audit tools that check for missing metadata, empty blocks, broken links, and orphaned content. An agent can run an audit and apply fixes in the same session.
Trigger and monitor builds
A build is a single tool call. The agent can trigger a build, poll for completion, check the output, and report errors — without a human touching the admin UI.
Connect an assistant
Complete setup, theming, and API documentation lives at docs.chinaski.net.
Read the documentation