For agencies, independents & in-house teams
The operating system for client work.
Every client, brand, and campaign in one place. Per-client AI drafts from each client’s own context — and nothing reaches the client without a human sign-off.
Built, not promised
A full operating layer — not another prompt box.
Model choice, production formats, storage, and accountable review work as one system. These are shipping product surfaces, not roadmap arithmetic.
One governed path
From model choice to approved, stored work
17
direct AI providers
8
deliverable formats
Required
enforced human gate
6
storage backends
Provider choice never bypasses the review gate. Every deliverable leaves with a recorded decision.
How it works
Brief in. Options out. A human signs.
- 01
Describe the job in chat — Tempera shapes it into a structured brief.
- 02
The client's agent drafts options from that client's own approved and rejected history.
- 03
A named owner approves before anything is client-ready — and, with a second person active, whoever ran it can't.
How client context becomes work
One client- Brand rules
- Approved work
- Review history
- The live brief
Client-ready work
Grounded in the right context and cleared by a named human.
Source + verdict attached
Real deliverables
Eight deliverable formats, from copy to storyboards.
Work lands as the file your client expects — with live formulas in the spreadsheets and editable text layers in the PSDs.
Illustration of the eight deliverable formats as file tiles — copy, documents, decks, social posts, graphics, layered graphics, spreadsheets, and storyboards — each with its delivery file format.
Plus the Media & AV pack: channel strategy, flighting plans, scripts, treatments, storyboards.
Who runs Tempera
The same platform, weighted for how you work.
One governed workflow — per-client context, drafted options, and a human gate. A workspace mode relabels the ceremony to fit your desk; it never removes the gate.
Agencies
“Every client in one place — and nothing ships without a named sign-off.”
The flagship: Client → Brand → Campaign → Activity, a dedicated agent per client, and account-team review.
Client
Brand
Campaign
Activity
Independents & freelancers
“My clients get agency discipline — without the agency overhead.”
Your own desk on the same model: every client in its own lane, and a review step that keeps your judgement in the loop.
For your own clientsIn-house brand teams
“One brand, every campaign — and legal can read the trail.”
Departments stand where clients stood, grounded in the brand book, with a flat monthly plan for the whole team.
For your brand teamAlso for consultants, studios of one, marketing teams, and content creators — the Solo desk fits any owner-operator. Creators
AI, engineered
Value from AI is an engineering discipline.
Not a bigger model and a bigger bill — the right model, the right context, and a meter on every call.
- The right model for each job
- Tasks route per client, agent, or task across 17 direct providers — a headline doesn't need a frontier model, and a storyboard shouldn't run on a cheap one.
- Only the context that matters
- Each client's agent reads that client's material — deduplicated, quote-stripped, scoped to the job — not a dumping ground of everything ever uploaded.
- Every call metered and capped
- Spend is attributed to the client, campaign, user, and model, with monthly budget caps that stop it before it happens — and bring-your-own-keys runs at provider cost.
Governed by default
Enforced, not promised.
Working solo, you approve your own work — recorded and audited. Add a second active user and the server blocks self-approval automatically.
See the controlsA human gate, enforced in code
With a second person active, whoever ran it can’t approve it.
Access scoped per client
Capability-based roles and per-user client scope, enforced on every read and write.
Your files, your storage
Six storage backends, including the disk you already run — the database keeps references, never bytes.
No model lock-in
Your keys, your providers. When a better model ships, you repoint the routing — no migration, no new tool, nothing re-learned.
The compounding advantage
Every verdict sharpens the next draft.
Models are replaceable. A client’s approval history is not. Tempera keeps each verdict attached, so the next draft starts closer to yes.
A cycle: a draft is reviewed, its verdict and reason return to that client’s memory, and the next draft starts closer to approval.
Review on web or phone. The verdict and its reason return to that client’s memory.
One core, every surface
One workflow, wherever the decision happens.
The web workspace, public API, and mobile companion share permissions, tenant isolation, budgets, context, and the same human review gate.
- Live today
The web workspace
The full workflow — intake, generation, review, delivery — on your agency’s own subdomain, white-labelled and multilingual.
- Live · opt-in
Public API v1
Agencies decide exactly which clients and capabilities each integration can reach — scoped, revocable keys, per-key rate limits, OAuth, signed webhooks, and OpenAPI 3.1 references.
- Live today
The mobile app
The review gate in your pocket: your “waiting on you” queue, approve or send back with notes, browse work offline, and pair with a QR scan.
One service layer underneath — an approval made on a phone is the same record as an approval made at a desk.
Why Tempera
Every claim above has a page of proof.
Grounded in each client's own context. Governed before anything ships. Priced for the clients you serve.
A platform that knows each client
Per-client agents trained on brand guidelines, approved and rejected work — producing copy, briefs, research, graphics, media plans, and storyboards.
Explore the platformGovernance built into the loop
A maker/checker review gate, capability-based access, encrypted credentials, and an append-only audit trail — enforced in code, not policy documents.
See the controlsPricing that follows the work
A platform fee per active client — a client with no AI work that month costs nothing. Start free, upgrade when the workflow proves itself.
See pricingQuick answers
All questions- Do we need AI keys to try it?
- No — mock mode runs the entire loop on labelled sample outputs, so you can judge the workflow before connecting a model.
- What does it cost?
- On agency plans, the platform fee follows active clients — dormant clients cost nothing, and seats are unlimited.
- Is our client data used to train shared models?
- No. Each client's agent draws on that client's own context and review history — nothing is fine-tuned into a shared model.
- Can we keep our own models and storage?
- Yes. Bring your provider keys, route each task to the model you choose, and point Tempera at your existing file storage — hosted or self-hosted.
- What happens before work reaches a client?
- A named human reviewer records the verdict. On team plans, the person who ran the task cannot approve it; Tempera enforces that separation on the server.
Before you commit
Evaluate the things that matter.
01
Full workflow, no AI keys
02
Dormant clients cost $0
03
Client data stays isolated
Test the loop, price it against active work, and verify the client boundary before you deploy.
Prove it on one client
Put one client's history to work. See the difference before you commit.
Run a real brief through Tempera in mock mode — from grounded options to a named human verdict and client-ready work.
No AI keys. No card. Nothing reaches a client without review.