Post to social media from Claude Code

Two routes into Claude Code: a server or a command

Postnavo is launching with both. The MCP server takes one claude mcp add command and a browser sign-in. The CLI is a program Claude Code runs in your shell, and it also works in scripts and CI, where there is no conversation.

There is a third way in if you already use Claude in a browser: connectors added on claude.ai are “automatically available in Claude Code” when you have logged in with a claude.ai account. On Team and Enterprise plans only admins can add those.[1]

Scopes decide who gets the server

A server added without a scope is local: it loads only in the project where you added it and is stored in ~/.claude.json. User scope makes it available in all your projects. Project scope stores it in .mcp.json at the project root, for version control.[1]

Claude Code prompts for approval in interactive sessions before it uses a project-scoped server from .mcp.json. In claude -p runs and cloud sessions it cannot show that prompt, and loads the server without asking.[1]

Organizations can restrict servers with managed settings: an allowlist and a denylist.[1]

Permission prompts for commands and MCP tools

In the default mode, labeled Manual, Claude Code asks before it runs a shell command, apart from a built-in set of read-only commands. Choosing “Yes, and don’t ask again” saves a rule for that command in .claude/settings.local.json.[3]

MCP tools take rules too. In an allow rule, mcp__<server>__* matches every tool of one server you configured. Anthropic notes that permission rules “are enforced by Claude Code, not by the model”.[3]

So you choose how much to hand over: approve each post by hand, or allow the Postnavo tools once. The approval setting Postnavo is launching with still holds posts for review when it is on.

Where Claude Code looks for skills

A skill is a folder with a SKILL.md file. Personal skills live in ~/.claude/skills/<skill-name>/ and project skills in .claude/skills/<skill-name>/, which a team can commit. Claude uses a skill when it is relevant, or you invoke it with /skill-name.[2]

Anthropic says Claude Code skills follow the Agent Skills open standard.[2]

Postnavo is launching with an installable agent skill that tells a coding agent how and when to call the CLI. The install command above is illustrative.

Announcing a release when nobody is at the keyboard

/loop re-runs a prompt on an interval, but tasks only fire while Claude Code is running, and recurring tasks expire seven days after they are created.[4]

Routines run on Anthropic’s cloud on a schedule, an API call or a GitHub event such as a release. They are a research preview on Pro, Max, Team and Enterprise, and the shortest schedule is one hour.[5]

A routine runs without stopping for approval, and can use every tool of an included connector, “including writes”. Servers you added with claude mcp add live on your machine, so a routine does not see them: Anthropic says to add the server as a connector on claude.ai.[5]

A routine that announces every release is the case Postnavo’s approval setting is launching for: the agent drafts and schedules, and a person approves.