Give your AI agents a design department.
Your agents work with everything: ideas, drafts, docs, PDFs, links, images. Canvora is the tool call that turns any of it into any type of visual: on-brand, across 100+ formats, generated natively in 150+ languages. One MCP connection or one API key.
Free credits, no card. Works with any MCP client or anything that can make an HTTP request.
Real session: an agent turning a brief into a finished visual set.
Raw image APIs give your agent pixels.
Canvora gives it deliverables.
An agent calling a diffusion model gets one pretty square with no memory of your brand. Here's what it gets instead:
Brand memory
Brand kits persist colors, fonts, logo, and tone per client. Your agent never has to re-describe a brand. It names one, and every output matches. The memory agents don't have, built in.
100+ real formats
Instagram post, LinkedIn carousel, pitch deck, email header, story, ad. Each has correct dimensions and safe zones. Your agent asks for a format by name instead of guessing pixel sizes.
Multi-slide deliverables
Carousels, presentations, and documents as coherent sets, consistent in style across slides and exportable to PDF, PPTX, or ZIP. Not nine disconnected images.
Edit, localize, export
The agent can revise its own work: "make the headline bigger", generate variations, localize a finished set into another language, and pull download URLs, all as tool calls.
Connect in two minutes
One remote MCP server. OAuth sign-in. No installation, nothing to host.
Claude
- 1. Settings → Connectors → Add custom connector
- 2. URL:
https://api.canvora.ai/mcp - 3. OAuth Client ID:
claude_mcp - 4. Sign in, then: "Use Canvora to create a LinkedIn carousel about…"
ChatGPT
- 1. Settings → Apps → enable Developer mode → Create app
- 2. MCP Server URL:
https://api.canvora.ai/mcp - 3. OAuth Client ID:
chatgpt_mcp - 4. Sign in, then + → Canvora in any chat
REST API (OpenClaw, Hermes, cron jobs, your app)
curl -X POST https://api.canvora.ai/api/generations \
-H "X-API-Key: vd_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"input_type": "text",
"input_content": "5 tips for remote work",
"output_formats": ["linkedin_carousel"],
"carousel_slides": {"linkedin_carousel": 5}
}'Keys live at Settings → API Access. Scoped, revocable, usage-tracked. Full API guide →
Agent frameworks like OpenClaw and Hermes connect through either lane: point their MCP client at the server URL, or give a skill the REST key. Step-by-step guide →
16 tools your agent discovers on connect
Full schemas over MCP. Your agent reads them and knows what to do.
chatDescribe what you need; Canvora plans formats, style, and content
confirm_generationApprove the plan and start generating
get_generationPoll status and fetch output URLs
list_formatsAll 100+ formats with dimensions
list_brandsThe account's brand kits
list_workspacesTeam workspaces the account belongs to
edit_imageRevise any output with an instruction
create_variationAlternate takes on an output
localize_generationRe-render a set in another language
convert_imageConvert an output between formats
upload_photoProvide photos to embed in visuals
export_visualsPDF, PPTX, or ZIP export
download_visualDirect file URLs for any output
get_edit_historyAudit trail of every edit
revert_editRoll back an edit that missed
check_creditsBalance check before big jobs
Built to be called by software
Predictable costs
Flat credit pricing an agent can budget: 10 credits per visual, 15 per carousel or deck slide. check_credits before a job, exact deduction after. No surprise token math.
Human in the loop by design
The chat then confirm flow means generation only starts on an explicit confirm call, and every output lands in your Canvora dashboard where a human can review, edit, or export.
Scoped, revocable keys
API keys are hashed at rest, revocable in one click, and tracked: per-key request counts and credit usage, so you know exactly what each agent spends.
Failures refund themselves
If a generation fails mid-run, credits are automatically refunded to their source. Agents retry safely without leaking spend.
Outputs on a CDN
Every visual gets a stable CDN URL your agent can post, embed, or hand to a scheduler. No base64 blobs to shuttle around.
Works on every plan
API and MCP access on all plans, including Free. Burst month? Campaign Packs add credits without a subscription, built for exactly the campaign-shaped work agents do.
Questions agents' humans ask
How does billing work when my agent generates? +
Same credits as the app: 10 per single visual, 15 per carousel or presentation slide, deducted from the connected account. Your agent can call check_credits before starting, and per-key usage tracking shows exactly what each agent consumed. Campaign Packs ($10 for 450 credits, 60-day validity) cover burst workloads without a subscription.
Does the agent generate without asking me? +
Generation requires an explicit confirm step (the confirm_generation tool), so a well-behaved agent proposes a plan first. Everything it creates appears in your Canvora dashboard for review, editing, and export, so the human stays in the loop.
Which agents and frameworks are supported? +
Anything that speaks MCP connects to the remote server with OAuth: Claude, ChatGPT, Cursor, Claude Code, and MCP-capable frameworks. Anything that can make an HTTP request (OpenClaw skills, Hermes, n8n, cron jobs, your own backend) uses the REST API with an X-API-Key header.
Is it safe to give an agent a key? +
Keys are stored as SHA-256 hashes, can be revoked instantly from Settings → API Access, and rate limits apply per key. A leaked or misbehaving key spends at most your credit balance. It can never touch billing or account settings, which require full login.
Your agent's next tool call is a design department.
Create a free account, connect the MCP server or grab an API key, and let your agent ship its first on-brand visual today.