MCP server for marketing, run it from Claude

An MCP server for marketing lets an agent do the marketing work, not just talk about it. Oetzi exposes every capability it has over MCP at /api/mcp: listening, scoring, drafting replies and posts, publishing, share of model checks, the people graph, and the outcomes ledger. Connect it to Claude Code, Claude Desktop, Cursor, or any MCP client and the agent works inside your workspace under the same rules you do.

What can Claude do with an MCP server for social media posting?

Anything you can do in the portal. There is one registry of capabilities and every surface renders from it, so the agent is never working with a cut-down version. It can read this morning's inbox and tell you which three conversations are worth a reply. It can draft the reply in your voice, hold it for your approval, and post it through Postiz once you say yes. It can run a share of model check, pull the engagers from a competitor's post, and write today's post on the cadence your posting plan set.

The difference from a posting MCP is what comes before the post. A scheduler's MCP publishes what your agent hands it. Oetzi's MCP also does the listening and the deciding, so the agent has a reason to post and a source to point at. The for agents page is the entry point, and the MCP docs cover connection and the tool list.

How do I use a marketing skill in Claude Code?

Install the SKILL.md. Oetzi ships one for Claude Code and OpenClaw that teaches the agent the capabilities, the credit rules, and the approval gate, so you can say "check what people said about us this week and draft replies" and it knows which tools to call. The Claude Code integration covers setup. The same registry backs a CLI and a REST API at POST /api/v1/capabilities/{name}, documented in the API docs and the CLI docs, for scripts and cron jobs that do not need a chat window.

How do I keep an agent from spending my credits?

Two mechanisms, both enforced on the server. First, keys carry scopes: read, write, spend, and post. A key with read scope can look at the inbox and the ledger and nothing else. A key without post scope cannot publish, whatever the agent intends.

Second, every paid call returns a quote first. The agent sees the price in credits before anything runs. The call only executes when it is resent with approved set to true and a max_credits cap. If the work would cost more than the cap, it does not start. This is the same gate a human clicks through in the portal, so an agent cannot find a cheaper path around it.

Reading is free over every surface. The API, the MCP server, events, and webhooks cost nothing. You pay for the same things a human pays for, sweeps, drafts, renders, and checks, at the same prices on the pricing page.

What happens with long-running work?

A sweep across seven platforms or a video render does not finish in one tool call. Those capabilities return a job. The agent can wait on the job, check it later, or move on and get told when it finishes. Webhooks and an event stream carry the results, so an agent that went to sleep still hears that the job finished, a new match landed, or a reply was approved.

Why does a marketing agent need its own MCP server?

Because an agent with a generic browser tool and a posting API has no memory of your market. It cannot tell a buyer from a bystander, does not know what you said last week, and has never heard your voice. Oetzi carries all of that, the voice doc, the people graph, the intent scores, the posting plan, and the MCP hands it to whichever agent you point at it. Honest note: if you only want a pipe to publish through, Blotato and similar tools exist and do that well. See Oetzi vs Blotato for the comparison.

Questions

Does Oetzi have an MCP server?

Yes, at /api/mcp. Every capability in the product is available through it, with the same scopes, prices, and approval rules as the portal. The same registry also renders a REST API, a CLI, and a SKILL.md for Claude Code and OpenClaw, so the agent can reach Oetzi from whichever surface fits the job.

Can Claude post to social media through Oetzi?

Yes, when the key carries post scope and you approve the draft. Claude drafts the post in your voice, the capability returns a quote, and publishing runs only after approval. Posts go out through Postiz to its 30+ platforms. Publishing itself is free, and a drafting run is one credit.

How does the MCP server stop an agent from overspending?

Every paid call returns a quote before it runs and executes only with approved set to true and a max_credits cap. Keys are scoped to read, write, spend, or post, so a key can be limited to reading. Reads, API access, events, and webhooks are free, so an agent watching your inbox costs nothing.

Is the marketing skill for Claude Code free?

Yes. The SKILL.md, the MCP server, the REST API, the CLI, webhooks, and events are all free. You pay only for metered work such as sweeps, drafts, renders, and share of model checks, at the prices on the pricing page. A free account starts with 10 credits and no card.