Plug, the navy puzzle-piece robot, pushes a glowing amber cable into a brass junction box that gathers four lit lines from a slotted panel, under a brass desk lamp.

Agents You can start this yourself today.

Connect Claude or ChatGPT to your brand guidelines.

Sign in with Jinn from Claude, ChatGPT, Claude Code or Codex, or make a token for Cursor and other clients, then ask it about your brand.

Start here

Connect your agent.

Sign in

Add Jinn from Claude's connector directory and sign in, with no token to paste. ChatGPT, Claude Code and Codex can sign in with Jinn too.

Or make a token

Make one in your account panel and keep it in your environment as JINN_MCP_TOKEN, never in a command or a shared file.

Add the server

One command for Claude Code, a config block for any other client, or one click for Cursor. The lines are below.

Ask

Ask your agent to call get_token_context first. It lists the brands it can reach. Then ask about your brand.

ChatGPT

Add Jinn to ChatGPT.

Add the address

Add Jinn as a connector by its address, https://app.jinn.works/api/mcp, with no header.

Sign in

Then sign in with Jinn and pick a brand.

Plugin

Jinn as a plugin.

For Claude Code, Jinn also comes as a plugin called Jinn Connect. It connects Claude Code to your brand record in one step, so Claude can read your voice, your kit and your design system while it works.

What it adds.

The Jinn connection, and one skill that checks the connection and tells you what your plan includes.

By hand, today.

Unzip the folder, run two commands, and paste your Jinn token when Claude Code asks. The token stays in Claude Code's secret store and goes only to Jinn. Ask us for it

In a plugin store.

Coming.

Any AI agent signs in with Jinn and reads your brand's record. You sign in with Jinn from your assistant, and any change it suggests to that record waits for a person's approval.

One call, start to finish

  1. Plug turning the key and the terminal lighting up

    The handshake first

    Plug makes the connection.

    The agent opens the connection with initialize and your token.

  2. Ocho setting the tool tags down in a row, one arm after another

    The tools, by name

    Ocho lays out the tools.

    tools/list returns the tools your token can see.

  3. A card sliding out of the terminal and Drift checking it against the blueprint

    The call and its answer

    Drift reads the answer.

    tools/call runs one tool, and your record answers.

The connection

Three ways in, one gateway.

Plug, Head of Connections

The gateway lives at https://app.jinn.works/api/mcp and speaks standard MCP (Model Context Protocol) over HTTP: JSON-RPC 2.0, with a token on every call and three methods, initialize, tools/list and tools/call.

MCP

One command. Keep the token in your environment as JINN_MCP_TOKEN, never in the command itself.

claude mcp add --transport http jinn https://app.jinn.works/api/mcp --header 'Authorization: Bearer ${JINN_MCP_TOKEN}'

Another client: the raw config block. It reads the token from JINN_MCP_TOKEN; if your client can't expand environment variables, put your token in the file on your own machine.

{
  "mcpServers": {
    "jinn": {
      "type": "http",
      "url": "https://app.jinn.works/api/mcp",
      "headers": {
        "Authorization": "Bearer ${JINN_MCP_TOKEN}"
      }
    }
  }
}

One-click install. Set JINN_MCP_TOKEN first; the link carries only a reference to it, never a secret.

Add Jinn to Cursor

Skills

Open-source marketing skills, free to use. Each one stands on its own, and a connected brand makes its work specific to you. Add one with a single line:

npx skills add JinnWorks/jinn-skills --skill brand-voice-content

Browse the catalog on GitHub

API

No client at all: POST to the gateway with Authorization: Bearer YOUR_JINN_MCP_TOKEN. The three methods, as wire bodies:

{ "jsonrpc": "2.0", "id": 1, "method": "initialize",
  "params": { "protocolVersion": "2025-03-26", "capabilities": {} } }
{ "jsonrpc": "2.0", "id": 2, "method": "tools/list" }
{ "jsonrpc": "2.0", "id": 3, "method": "tools/call",
  "params": { "name": "get_brand_dna_public", "arguments": { "slug": "paleo-pro" } } }
The connection, in one table
WhatHow it works
LimitsThe agent connection's limits: 120 requests a minute per token; a rotated token's old secret keeps working for 24 hours by default and 48 hours at most; a demo token lasts 60 minutes, with 3 active per address and 5 new ones an hour.
TokensA token is shown once. Jinn keeps only a one-way fingerprint of it.
The activity logThe activity log records the tool, the brand, the time and a fingerprint of the request, never the request itself. It is cleared after 90 days.

The tools

The tools, by name.

Your token's own tools/list returns the tools it can see. A tool above your plan answers TIER_REQUIRED with the plan it needs; a tool that isn't offered to you reads as an unknown tool, on purpose. The live list, with each tool's inputs and answers, is in the developer reference.

ping

Checks the connection end to end. No arguments.

get_token_context

What your token can reach: your brands, its scopes, its plan and when it ends. Call it first.

get_brand_dna_public

The public part of a brand record: identity, voice, positioning, messaging pillars. What a demo token reads. On your own brand, Connected plan and up.

get_brand_design_md

Your brand's DESIGN.md, a design system written for agents to read. Connected plan and up.

get_brand_design_tokens

Your design tokens as DTCG JSON (the open design-token format): colors, type, spacing, radius, motion. Connected plan and up.

ask_brand

Ask a plain question about your brand; get back the facts in your record that bear on it, and what the record can't answer. The question is logged for the brand's owner. Connected plan and up.

get_brand_kit

The render-ready kit: colors, fonts, logo, name, spacing. Brand plan and up.

get_brand_dna

Your full brand record: identity, story, positioning, audience, voice, messaging pillars. Brand plan and up.

get_brand_products

Your products, one by one: description, ingredients, certifications, dimensions, link. Brand plan and up.

get_brand_design_system

The design system measured from your live site, each value with how sure it is. Brand plan and up.

list_design_extraction_runs

Each time your site's design was measured, newest first. Brand plan and up.

get_product_design_system

One product page's design, as the ways it differs from your brand's. Brand plan and up.

create_brand

Starts a brand of your own from its name and website, with a free preview read from the homepage. Not on a demo token.

get_brand_status

Where a new brand stands while Jinn reads it: queued, reading, ready or failed. Not on a demo token.

list_my_brands

Every brand you own on Jinn, and which one your agents read now. For agents signed in with Jinn.

switch_brand

Changes which of your brands your agents read, from the next call. For agents signed in with Jinn.

propose_brand_edit

Proposes changes to named fields of your record. Each one waits in your inbox, and nothing changes until you approve it. Connected plan and up.

review_asset

Reviews one of your images against your kit and voice: a verdict, findings by area and a short summary. Capped per person per day. Connected plan and up.

search_assets

Searches your own published assets by a short query. Each result has a download link that works for an hour. Connected plan and up.

brief_task

Before a task, a brief built from your record: the voice to use, the design rules, and what the record doesn't know; from Brand up, the products involved. Connected plan and up.

get_brand_skill

Packs your brand as a skill folder your agent installs: voice and design references, each file dated. Connected plan and up.

Tokens

A token's life, start to finish.

Make

Make a token in your account panel. The secret is shown once and stored only as a hash.

Rotate

Rotating gives you a new secret while the old one keeps working for 24 hours by default and 48 at most, so nothing breaks mid-swap.

Revoke

Revoking takes effect at once, and a revoked token can't be rotated back to life.

Demo

A demo token lasts 60 minutes and can only read the showcase brands. Each IP address can hold 3 at once and make 5 an hour.

Demo token

Try the gateway first.

No account yet: ask for a demo token from the command line. The answer carries the token, the address to connect to and how to add it.

curl -X POST https://app.jinn.works/api/agents/request-demo-token

Entitlement

Your plan is read on every call.

A token carries no frozen plan. Its access is worked out from your live subscription on every request, and refused if the subscription can't be read, so an upgrade, a renewal or a lapse takes effect on the next call. There is nothing to re-make.

Errors

Limits and errors, as your agent sees them.

What each error means
CodeWhat it means
RATE_LIMITED (-32003)More than 120 requests a minute on one token. The response carries a Retry-After header to wait out.
TIER_REQUIRED (-32004)A tool above your plan. The error carries the plan it needs and an upgrade link.
QUOTA_EXCEEDED (-32005)Your plan's monthly allowance is used up.
METHOD_NOT_FOUND (-32601)A tool that isn't offered to you reads as an unknown tool, so its existence stays hidden.
FORBIDDEN_SCOPE (-32002)The token doesn't carry the permission this tool needs.
not_foundAny brand you can't reach returns the same not-found as a brand that doesn't exist.
token_malformedThe header isn't a well-formed Jinn token. Fix it rather than retry.
token_invalidNo live token matches: the secret is wrong or was never issued.
token_revokedThe token was revoked. Make a new one in your account panel.
token_expiredThe token or its plan lapsed. The error carries a renewal link.

The four token errors answer HTTP 401 with the code in error.data.code. A tool error (-32000) also carries a machine-readable data.kind and a retryable flag, so an agent can decide whether to retry, fix or stop without reading the message.

Reference

Terms and the full reference.

The API terms govern gateway use. Each tool's request and response shapes live in the developer reference.

See What Each Plan Adds

Bristle painting a jar, Ocho sorting folders, Wire on the phone and Proof checking a page, each in a lit arched window at night.

Stuck on a step? Tell us where.

  1. Tell us about your brand.
  2. A person on our team reads it.
  3. We read every application and reply within two business days.