Platform — MCP + API
One endpoint, every agent that makes.
Query your brand profile over MCP or REST. Any agent, any tool, one source of truth. Read before it writes a word or makes a pixel.
Built for agents
Every tool reads the same source before it makes anything.
Aravi speaks two protocols: Model Context Protocol for agentic tools that support it, and a plain REST API for everything else. Same profile, same rules, same answer — no matter which one your stack calls.
mcp · brand-profile
Request
GET /mcp/brand-profile agent: "figma-make" scope: voice, palette, logo
Response · 200 · 84ms
{
"voice": ["confident","direct","no jargon"],
"palette": {"ink":"#17100d","teal":"#45848d"},
"type": {"primary":"GT America"},
"profile_version": "2.3"
}
Design agents
reads: palette, type, logo
Copy agents
reads: voice, terms
Code agents
reads: tokens, radii
Image agents
reads: imagery rules
1
Generate a scoped key
Create an API key from your dashboard, scoped to read-only or read plus validate.
2
Point your agent at the endpoint
One call, either protocol. The response is identical whether it comes over MCP or REST.
GET https://api.aravi.ai/v1/brand-profile Authorization: Bearer <key>
3
Read before you make
The agent gets voice, palette, type, and imagery rules back before it generates anything — on-brand by default, not by review.
GET
/v1/brand-profile
Fetch the current profile
POST
/v1/validate
Score a draft against the profile
POST
/v1/log
Log an approve or reject — feeds Compound
GET
/mcp/brand-profile
MCP-native read, same data
MCP
For agentic tools that speak Model Context Protocol natively. Connect once, and every session already knows your brand — no re-prompting, no pasted style guides.
REST API
For everything else. A plain HTTP endpoint any pipeline, script, or custom agent can call, with the same fields and the same answer.
Ready to wire it in?
Join the first group of brands giving every agent they touch a source of truth to query.
Join the waitlist