AAravi
MCP & API

Developers

Connect Aravi to Claude, ChatGPT, Cursor and your own agents over MCP — one integration, and every AI tool can read your brand, create on-brand, and check its work before publishing.

Your MCP server

One URL, every client. Interactive apps sign in with OAuth — no key needed. Headless agents and servers authenticate with an API key.

https://aravi-api-production.up.railway.app/api/mcp
Streamable HTTPOAuth 2.1 + PKCEDynamic client registrationAPI keys
Works with Claude ChatGPT Cursor Claude Code VS Code

Connect your client

Pick your tool, then your auth method — setup takes about a minute.

Auth method
Recommended — nothing to manage
1
In claude.ai, open Settings → Connectors and click Add custom connector.
2
Paste the MCP server URL:
https://aravi-api-production.up.railway.app/api/mcp
3
Click Connect— a browser tab opens Aravi's sign-in and authorizes automatically. No key needed.

Know it worked: after connecting, ask your assistant “List my Aravi brands” — you should see your workspace brands come back. Then try a starter prompt below.

Try these first

Copy a prompt into your connected assistant — each one exercises a different part of the brand layer.

What agents can do

The full tool surface agents get once connected.

Understand

aravi_get_brand_context

Identity as a ready-to-inject text block — for on-brand copy and prose

Aravi
aravi_get_brand_contract

Structured JSON contract — every field with confidence + provenance + version

New
aravi_get_brand_design_md

DESIGN.md design tokens for generating on-brand UI code

Aravi
aravi_get_brand

Full four-layer identity (strategic, verbal, visual, imagery)

aravi_get_brand_audit

Maturity score, missing signals, prioritized fixes

aravi_get_brand_assets

Canonical logo, style anchors, fonts, palette — real assets, not descriptions

New
aravi_list_brands

All workspace brands with status and key facts

aravi_search_library

Semantic search over the brand image library

Create

aravi_create_brand

Onboard a brand from a website URL or Instagram handle (~60–90s)

aravi_generate_image

On-brand image generation — identity, anchors and logo injected automatically

aravi_get_image

Collect a generation (returns brand-match score with the URL)

aravi_list_images

Browse generated images

Guardrail

aravi_check_brand_fit

Score any image or copy 0–100 against the brand — violations + fixes before you publish

New

Account

aravi_get_account

Authenticated user, workspace and plan

aravi_check_credits

Remaining generation credits

API keys

For headless agents, CI, and server-to-server calls. Keys are workspace-scoped and shown only once.

No API keys yet — interactive clients can use OAuth without one.

Brand-change webhook

Get a signed POST whenever a brand's identity changes — so your agents and caches invalidate instead of polling.

Payload: {event, brand_id, identity_version, updated_at} — HMAC-SHA256 signed via the x-aravi-signature header. HTTPS only; leave empty to disable.

How agents should use Aravi

This guidance is also served to agents automatically via MCP server instructions.

1. Find or onboard the brand — aravi_list_brands, or aravi_create_brand with a website URL or Instagram handle (~60–90s, poll with wait=true).

2. Read the layer — get_brand_context for copy, get_brand_design_md for UI code, get_brand_contract for structured data with confidence scores.

3. Create — describe the scene and intent; Aravi injects the identity, style anchors, and logo automatically. Collect with aravi_get_image — the result includes its brand-match score.

4. Guard before publishing — run aravi_check_brand_fit on anything made outside Aravi. Score ≥ 70 with no hard violations = on-brand.

Each generation costs 1 credit. Connected clients share your workspace's balance.