Running Multiple Client Brands in One Workspace

You run multiple client brands in one SocialPower workspace by creating one brand per client. Each brand carries its own Brand Brain and its own connected accounts. A generation for one client reads only that client's record. Team roles control who sees what, and a client can be given view-only access. Your plan decides how many brands you get.

The failure this prevents is the one every agency knows. A caption written for a skincare D2C brand goes out on a fintech handle. A banned word slips through because the person writing it was working from memory at 11pm. Separation has to live in the tool, not in your team's discipline.

Three things are scoped to the brand, not to the workspace.

• The Brand Brain. Voice, audience, reading level, sentence length, always-say facts, banned words, emoji policy, hashtag ceiling, CTA style and claims policy. One record per client.
• Connected accounts. The Instagram, LinkedIn, X, YouTube and other handles you publish to are attached to a brand. They do not show up while you are working inside a different brand.
• Generation context. When you write a caption, the model is conditioned on that brand's record before a word is produced. It is not a chat that you have to keep reminding.

One thing is scoped to the workspace: your AI provider. You pick OpenAI, Claude, Gemini, Groq, Mistral, DeepSeek or HuggingFace once, and every brand uses it. You cannot run Claude for one client and Gemini for another today. If the key is missing or misconfigured, a deterministic mock provider answers instead of the request hard-failing, which keeps a bad key from taking your whole queue down mid-campaign.

Most agencies keep client voice in three places: a PDF brand guide nobody opens, a Notion page that is four months stale, and the head of content's head. The Brand Brain is the same information written as fields a machine reads before it generates.

The fields that do the most work across a client roster are the restrictive ones. Banned words stop a compliance problem before it reaches a draft. A claims policy is the difference between "clinically proven" and "formulated with" for a beauty client. A hashtag ceiling stops a junior from stacking thirty tags on a LinkedIn post. Always-say facts carry the things a client will email you about if they are missing, like a registration number, a founding year, or the correct spelling of a product name.

Fill these in during onboarding, once, with the client on the call. That hour is the highest-leverage hour of the engagement. Everything generated afterwards inherits it, including work done by a freelancer who has never spoken to the client.

You connect a client's handles inside that client's brand. Publishing targets Instagram, Facebook, LinkedIn, X, Pinterest, Threads, TikTok and YouTube, each at its own native dimensions and character limits, so the same idea comes out sized correctly per network rather than cropped by hand.

Practically, this means the wrong-account mistake requires you to be inside the wrong brand entirely, not to mis-tick a checkbox in a list of forty handles. It is a coarser, safer boundary.

There are five roles. The ladder runs owner, admin, manager, editor, viewer.

The viewer role is the one that changes client relationships. Instead of exporting a content calendar to a spreadsheet every Monday, you give the client a login and they watch the queue fill. They stop asking what is going out this week, because they can see it. Business plan includes 20 team members, which for most agencies means the internal team plus a viewer seat for each active client.

Brand count is the number that matters for an agency. Generations and account count rarely bind first.

Run the per-brand maths before you pick. Pro is roughly ₹500 per brand per month at three clients. Business is roughly ₹400 per brand at ten. If you are at four clients, you are on Business and paying for six empty slots, so the honest question is whether client five and six are close enough to signing to make the jump now.

All prices exclude 18% GST. Starter is ₹589 all-in, Pro is ₹1,769, Business is ₹4,719. Payment is UPI, card or net banking. Full plan limits are on the pricing page.

Autonomous mode is where agencies get nervous, and the design accounts for that in two ways.

First, the approval gate is on by default. Nothing publishes until a human releases it. An agency running client work should leave it on. Your manager or account lead becomes the release point, and the client's viewer seat means they can see what is queued behind that gate.

Second, the daily post cap is enforced in the worker, not just displayed in settings. The worker counts today's AI posts and stops at the plan's ceiling. That is a real hard stop, not an intention.

Plan the cap against your roster. Fifteen posts a day on Business across ten active clients averages 1.5 posts per client per day. That is enough for a normal always-on cadence and not enough to run ten campaign launches in the same week. If a client needs a heavy fortnight, budget the rest of the roster down for it.

One thing to be clear about: the monthly spend cap is stored and shown, but the worker does not enforce it. Only the daily post cap stops the worker. Watch provider spend yourself if you are on a metered key.

An agency evaluating this should know what it will still do elsewhere.

• Analytics readback, trend detection, SEO, competitor tracking and rendered video are on the roadmap and labelled SOON in the product. Five of thirteen agents are not built.
• There is no comment moderation and no inbox management. Community management stays in your existing tool or in the native apps.
• There are no paid ads. This is organic publishing only.
• Reporting to clients is still your job. Pull numbers from each platform's native insights until the analytics agent ships.

If a client is buying you for reporting and community management, SocialPower covers the production and scheduling half of that retainer, not all of it. Price accordingly.

Do it in this order. It takes about an hour per client and it is the whole setup.

1. 01Create the brand. Name it exactly as the client's legal or trading name appears, so nobody in your team guesses.
2. 02Fill the Brand Brain with the client on the call. Voice, audience, reading level, sentence length first.
3. 03Write the restrictive fields while they are still on the call: banned words, claims policy, emoji policy, hashtag ceiling. Ask directly what has caused a problem before. That question fills these fields faster than anything else.
4. 04Add always-say facts. Registration numbers, correct product spellings, the founder's name, the disclaimer legal insists on.
5. 05Connect the accounts for that brand only. Verify each handle after connecting.
6. 06Generate five captions and read them against the brief. Fix the Brand Brain, not the captions. If the output is off, the record is wrong.
7. 07Leave the approval gate on. Set the daily cap to what you can genuinely review.
8. 08Invite the client as a viewer and send them the link on the same call, so the first thing they see is a filled queue rather than an empty dashboard.

Step 6 is the one teams skip and the one that pays. A caption you fix by hand helps one post. A Brand Brain field you fix helps every post that client gets for the rest of the year.