For coding agents

Blog CLI for AI Agents

Blogizi is a CLI and hosted MCP server built for coding agents — Claude, Codex, Gemini, OpenClaw, Hermes, and anything that can run shell or call tools — to draft and publish markdown posts on your own Blogizi blog.

View Docs

Blogizi drafts by default — publishing live always requires explicit confirmation.

Install

Get Started in Seconds

Install the CLI, authenticate with your account API key, and push a draft from any markdown file.

Terminal

npm install -g blogizi
blogizi auth YOUR_API_KEY
blogizi draft ./posts/my-first-post.md

Prefer a skill install? npx skills add bytemindlab/blogizi-agent

Capabilities

Built for how agents actually ship posts

No invented platforms or vanity metrics — just what blogizi-agent and the hosted MCP do today.

AI Agent Ready

Commands and MCP tools return clear, structured results so coding agents can parse status, pick a project, and decide the next step without guessing.

Markdown & Frontmatter Native

Posts are plain .md files with YAML frontmatter — title, slug, description, and keyword — so agents edit what you already version in git.

Draft-First Workflow

Every post lands as a draft by default. Going live is a separate, explicit step — blogizi publish or MCP publish_post — so agents cannot ship without confirmation.

Obsidian Vault Integration

Publish straight from a vault note with the Obsidian plugin— same account API key as the CLI.

Multi-Project Support

One account key covers every blog. Use blogizi use <slug> (or projectSlug on MCP tools) so the agent posts to the right project.

SEO Metadata Built-in

Keyword, description, and slug are first-class frontmatter fields — not afterthoughts bolted on after publish.

Workflow

How It Works

Authenticate once. Let the agent write the file. You review the draft before anything goes live.

01
Authenticate & select a project
Copy your account API key from Dashboard → Account → API. Run blogizi auth, then blogizi use when you have more than one blog.
Auth
blogizi auth YOUR_API_KEY blogizi use my-app
02
Draft your post
The agent writes a .md file with Blogizi frontmatter, then runs blogizi draft. The post is saved as a draft — not live.
Draft
# Agent writes the .md with YAML frontmatter, then: blogizi draft ./posts/how-we-built-auth.md
03
Review & publish
You review the draft in the dashboard (or the local file). Publishing live always requires an explicit blogizi publish — agents should not treat draft as go-live.
Publish
# After you review the draft in the dashboard: blogizi publish ./posts/how-we-built-auth.md

Agent runtimes

What is OpenClaw / Hermes-Agent?

AI agent runtimes give a model a shell, tools, and skills so it can take real actions — not just chat.

OpenClaw

Open-source, skill-based, and often chat-app-native: you message the agent from apps you already use, and it runs skills on your machine. Blogizi plugs in as a skill via the bundled SKILL.md.

Hermes-Agent

A comparable skill-oriented agent runtime. Same idea for Blogizi: discover the skill, authenticate with your account API key, then draft and publish through the CLI (or MCP when shell is limited).

How Blogizi becomes a skill

Bundled SKILL.md is auto-discovered by compatible runtimes.
Install with npx skills add bytemindlab/blogizi-agent.
Works with any LLM backend the runtime supports — Claude, GPT, Gemini, and others.
Prefer CLI when a shell is available; fall back to hosted MCP in sandboxes (see below).

Skill install

# Install the Blogizi skill for agent runtimes
npx skills add bytemindlab/blogizi-agent

# Or use the CLI directly once authenticated
blogizi draft ./posts/my-first-post.md

Anthropic

Also Works with Claude Code / Claude Cowork

Same Blogizi account — pick CLI or MCP based on whether the agent has a real local shell and config.

Claude Code

Reads SKILL.md, writes the markdown with frontmatter, and runs blogizi draft / blogizi publish when you ask it to write and ship a post. Keep publish behind your explicit OK.

Claude Cowork / sandboxed agents

Cowork's sandbox often cannot share ~/.blogizi/config.json. Use the hosted MCP server instead of the CLI — same draft-first tools: list_projects, draft_post, update_post, publish_post.

MCP connector

claude mcp add blogizi --transport http \
  --url https://blogizi.com/api/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Full setup: /docs/mcp

Reference

Command Reference

Core CLI surface agents should know. See the repo README for update and edge cases.

Command	Description
`blogizi auth <apiKey> [--project <slug>]`	Save account API key (and optional project)
`blogizi use <projectSlug>`	Set active project for draft/publish
`blogizi draft <file>`	Save a local .md file as a draft
`blogizi publish <file>`	Publish a local .md file live

Also available: blogizi update <file> to re-push an existing slug without inventing a new one. MCP equivalents live at /docs/mcp.

Let agents draft. You decide what goes live.

Agents write the markdown. Blogizi handles frontmatter, SEO fields, hosting, and publishing — with drafts by default until you confirm.