How postmakr works.
From idea to published post: this page explains the system behind postmakr, including the MCP connection and the complete tool reference for your AI model.
What is postmakr?
postmakr is the content platform for your social media: your own AI creates carousels, reels and stories for Instagram and TikTok, and postmakr organizes, stores, schedules and posts them. In short: your AI creates, postmakr organizes and posts. The special part: postmakr doesn't ship its own AI model. Instead it works through the open Model Context Protocol (MCP) with the model you already use: Claude, GPT, Gemini or any other.
Every brand gets its own workspace. That's where its style guide, ideas, post archive and learnings live, everything your AI model needs to create posts that genuinely look like your brand. And the hardest rule in the system: nothing goes live without your approval.
The use cases show who postmakr is built for, and What makes postmakr different explains how it differs from scheduling tools and pure AI image generators. The key terms are explained in the glossary.
How the system works
postmakr thinks in two interlocking cycles. The production loop takes posts from idea to publication, and the learning loop pulls every post's numbers back into the system.
The production loop
- Ideas: your AI model suggests what you could post next, or you add an idea yourself. You decide: implement, keep or discard.
- Creation: your AI model builds a complete post from the implemented idea (copy, slides, rendering), grounded in your style guide and similar posts from the archive.
- Approval: you see the finished draft in the dashboard, adjust the caption if needed and approve it, or have your AI model revise it.
- Publish: the approved post moves into the archive, and at the scheduled time postmakr publishes it through your connected channels.
The learning loop
- Performance: views, likes, retention: every published post's metrics flow back in.
- Analytics: the numbers become patterns: which hooks pull? Where do people drop off?
- Learnings: the insights go to your AI model as rules for the next idea and the next creation. Every post makes the next one better.
The features in detail
Ideas
Ideas are where everything starts. An idea is a suggestion for what you could post next, with a working title and a short note on the angle. Ideas come from your AI model, which knows your brand and your learnings and saves its suggestions in your workspace, or from you, whenever something occurs to you.
In the dashboard you decide on every idea: Implement, Keep or Discard. If you implement an idea, your AI model builds the post. If you keep it, it waits in the list until the time is right. If you discard it, it's gone. That way nothing gets lost, and nothing is created that you haven't chosen.
Every format, built from your format catalog
Your AI model works against a format catalog: carousel, animated carousel, reel, story or sharepic. Each format brings its own spec for structure, slide count and dramaturgy. Your AI model doesn't freestyle; it builds along that spec so every post hits your brand's style.
Finished doesn't mean frozen: until approval you can have any draft revised by telling your AI model what should change. You can also adjust the caption by hand in the dashboard, restore earlier file versions, and rendering happens on the server, as MP4 or PNG, bundled as a ZIP to download.
Your own AI model
postmakr deliberately ships no AI model of its own. Instead, the model you already pay for (Claude, GPT, Gemini or any other) connects to your workspace through the open Model Context Protocol. That has three advantages: you pay no duplicate AI costs, your content runs through your own account, and you decide which model you trust.
postmakr stores no third-party API keys. The connection runs over a workspace-scoped access key that you generate in the dashboard and can revoke at any time. How the connection works technically is covered in Understanding MCP.
Team workspaces
Every brand lives in its own workspace with its own style guide, ideas, archive and learnings. Nothing mixes. An agency can run ten client brands in parallel without one brand's tone leaking into another's posts.
Within a workspace, several people work together: each access has its own key, produces with its own AI account, and only sees the workspaces that concern them. Approval remains a separate, clearly defined step in the dashboard, and you decide as a team who takes it on.
Performance learning
After publishing, the second half of the system kicks in. Every post's metrics (views, likes, comments, shares, drop-off points) flow back into the workspace and get condensed into patterns in the analytics: which hook types pull with your audience? Which formats perform? On which slide do people bail?
The patterns become learnings, and those land exactly where the next production starts: in the context your AI model receives on its next idea round and its next /create. The system gets a bit sharper with every published post, not as a black box, but readable in the dashboard.
Understanding MCP
The Model Context Protocol is an open standard that gives AI models a uniform way to talk to external systems, often described as "USB-C for AI applications". Instead of building a custom integration for every tool, your AI model connects once via MCP and can then use everything the server offers.
postmakr is such an MCP server. Your AI model (the MCP client, e.g. Claude in the terminal or in the app) connects to your workspace and gets three things:
- Tools: actions the model can perform: saving ideas, saving drafts, rendering slides, starting workflows. All 30 are listed in the reference below.
- Context: your brand's style guide, format catalog, archive and learnings, so the post sounds like you.
- Workflows: guided sequences like
start_ideasor/createthat walk the model through ideas and production step by step.
All trademarks and logos are the property of their respective owners. No partnership implied.
Tool reference
Your workspace exposes these 30 tools to the connected AI model via MCP. You don't need to memorize them, because your model picks them itself. But they show exactly what the system can do.
Context & archive
get_workspace_context | Returns the workspace overview: brand, style guide, format catalog and current learnings. |
update_brand | Updates the brand and style guide of the workspace, for example after your AI has read your website. |
update_workspace_context | Updates the background information your AI model should always keep in mind for this workspace. |
get_reference_posts | Finds the two or three most similar already-published posts for a topic, as a style template for the new one. |
archive_search | Searches the entire post archive by topic, hook or format. |
Ideas
save_idea | Saves an idea in the workspace: title, angle and a short rationale, so you can implement, keep or discard it. |
list_ideas | Lists the ideas in the workspace with their status, so nothing gets suggested twice. |
References
save_reference | Saves a post by someone else that you like as a reference, together with your AI's analysis and what to adopt from it. You add the images in the dashboard. |
list_references | Lists the references in the workspace, optionally filtered by keyword or format. |
get_reference | Fetches a reference with its analysis and up to six images, so your AI can look at it before creating. |
Drafts
list_drafts | Lists all current drafts with their files. |
save_draft | Saves a file into a draft: copy, slides, captions. |
read_draft | Reads a file from a draft, for example to revise it. |
Rendering
render_slide | Renders a slide to a finished PNG at post resolution. |
start_animation_render | Starts rendering an animated slide as a video (MP4). |
check_render_status | Checks the status of a running video render. |
Series formats
list_templates | Lists the saved templates for series formats with name, fields and source post. |
create_template | Saves an existing post as a template for a recurring series where only the date, issue or topic changes. |
create_from_template | Creates a new issue from a template as a draft and says exactly which spots need the new values. |
Sounds & footage
find_sounds | Finds fitting sounds for a reel by mood, sound or image. |
suggest_sounds | Suggests sounds for a finished draft. |
find_footage | Searches your footage library for fitting scenes. |
use_footage | Brings a clip from the library into a draft. |
check_reel_sound | Sound check for a finished reel: cuts, loudness and whether the voice matches the script. |
Workflows
start_onboarding | Starts the guided onboarding: your AI reads your website and creates brand and style guide. |
start_ideas | Starts the guided idea round: your AI suggests ideas, and you decide whether to implement, keep or discard them. |
start_create | Starts the guided creation of a post, optionally from a saved idea via idee=<id>, including format spec and references. |
start_revise | Starts revising an existing draft based on your feedback. |
start_approve | Starts the approval check: final quality control before publishing. |
approve_post | Archives the approved post and readies it for publishing. |
Security & data
postmakr is built for brands that take their content seriously. A few hard principles apply:
- Strictly separated workspaces: every brand lives in its own isolated data area. Nobody sees another brand's drafts, archive or numbers, not even by accident.
- One central login: sign-in runs through a dedicated, self-hosted identity service (Keycloak). Your account belongs to you, and passwords never touch the application itself.
- No third-party API keys: your AI model runs on your own account. postmakr stores no third-party keys, only the workspace access key that you generate yourself and can revoke at any time.
- Hosted in Germany: the platform runs on servers in Germany, with daily backups of all workspaces.
- Your content is yours: everything created in your workspace (posts, media, learnings) remains your property and can be exported at any time.
Frequently asked questions
Do I need my own AI subscription?
Yes. postmakr deliberately ships no AI model of its own and works with the subscription you already have (Claude, GPT, Gemini or any other). That way you pay no duplicate AI costs, and your content runs through your own account.
What can I do in the dashboard without AI?
Everything around review and publishing: you see every draft, adjust the caption by hand, approve, schedule the date in the calendar and have the post rendered as MP4 or PNG and downloaded as a ZIP. Rendering is done by the server, whether you trigger it in the dashboard or your AI model triggers it via a tool. Changes to the slides themselves are handled by your AI model.
Can my team work with me?
Yes. A workspace can have several accounts: each person works with their own login and their own AI account on the same brand, with clear approval paths.
Does postmakr publish directly to Instagram and TikTok?
You schedule approved posts in the calendar. If the channel is connected in your workspace, postmakr publishes the post automatically at the scheduled time. For channels that aren't connected, the entry stays a reminder and you upload the rendered post yourself. Either way, only what you have approved gets published.
What does postmakr cost during the beta?
Nothing. During the beta phase you test everything for free. Fair pricing comes afterwards, and early testers get special conditions.
Connect & get started
The connection takes just a few minutes to set up:
- 1. Access and workspace: you request beta access, and we set up your workspace together with you. Your AI model then reads your website and creates brand and style guide.
- 2. Connect your AI model: add the postmakr MCP server in your AI client, for example in Claude Code in the terminal or in Claude Desktop. Sign-in happens via browser login. Only clients without browser login need an API key, which you create in the dashboard under "Connection & API keys". All variants are on the MCP page.
- 3. Produce:
start_ideasbrings fresh ideas from your model, and/createhas your model build the post from the idea you implement. The final approval is always yours, in the dashboard.
A detailed step-by-step guide with examples is in Create posts with AI.