The reason AI copy sounds generic is that nobody told it otherwise
A model with no brand context defaults to the average of everything it has read. That average is exactly the voice you have spent years trying not to have. Tell your agent once, in its own instructions, and every post it sends through PostMCP carries it.
“🚀 Excited to announce a game-changing update that will revolutionise how teams collaborate! We’re thrilled to unveil our latest innovation. #innovation #growth #futureofwork”
“Workspaces shipped. One account, one bill, every client kept separate. It took longer than we said it would and we should have built it first.”
Six things to write down, and the reason each one earns its place
Half a page is enough. Put it wherever your agent keeps its standing instructions, and it applies to every draft the agent hands to PostMCP.
The single biggest lever on whether output reads as yours. Your agent carries it into every draft, so it governs rhythm as much as vocabulary.
Changes what gets explained and what gets assumed. The same fact written for engineers and for buyers is two different posts.
Your rules, in your words. This is where legal constraints, positioning boundaries and house style live.
The words you want reinforced, and by omission the words you have been trying to stamp out for two years.
Generated visuals inherit its palette and composition instead of drifting into whatever the model felt like today.
Real files your agent can attach as mediaUrl, described well enough that it picks the right one for the post it just wrote.
Every agent already has a place for it
One Project per brand; the PostMCP connector attached to each.
A custom GPT per client keeps voices and workspaceIds apart.
Lives in the repo beside the product it talks about.
The skill ships the tools; your brief rides alongside them.
One workspace, several voices
An agency does not have a brand voice — it has eleven of them. A founder-led company has at least two: the company page and the person. Give each its own Project or GPT, and its own PostMCP workspace, so the right voice reaches the right accounts and nothing else.
How agencies use thisBrand voice questions
- What is a brand voice AI?
- It is a written description of how your brand writes and looks, applied to every piece of AI-generated content. With PostMCP that description lives in the instructions of the agent you already use — voice, audience, rules, vocabulary, a style reference and the assets it may attach — and the agent reads it before it drafts a single line for create_post.
- Why keep the voice in my agent rather than in PostMCP?
- Because that is where the writing happens. Claude, ChatGPT and Cursor all have a standing place for instructions — a Project, custom instructions, a CLAUDE.md — that every conversation starts from. Put the voice there and it applies to your social posts, your changelog, your emails and anything else the same agent writes, instead of only to one form in one tool.
- How is a voice brief different from just prompting well?
- A good prompt works once. A brief in the agent’s standing instructions works every time, for every person using that Project, and for the agent running unattended at 6am. It also survives model changes — you are describing your brand, not tuning a specific model’s quirks.
- Can I have more than one brand voice?
- Yes, and most serious users do. One Claude Project or custom GPT per client, each with its own brief, and one PostMCP workspace per client so the right accounts are the only ones in reach. Every tool takes a workspaceId, so the agency GPT for Northwind cannot post to Halcyon.
- Does the brief affect generated images too?
- Yes. Tell your agent which reference image to pass as styleImageUrl to generate_image and the visual output follows it — palette, composition, mood. It is the difference between a set of images that look related and a set that obviously came from different prompts.
- Does everyone on the team share the same voice?
- Share the Project, the GPT or the repo and they do. Each member connects with their own PostMCP API key, so the agent acts as them — with their role’s permissions — while writing against the same brief. Consistency across a team is exactly what a documented voice is for.