Your assistant. Your artifacts. Your continuity.

Your AI.
Now with Receiz.

Create proof objects. Verify the files that matter. Build on an identity and history you can carry with you. Bring Receiz into the AI you already use.

Choose your setup Discover Receiz Mind

Manual installation is available independently of marketplace approval.
Tool execution requires a compatible MCP host.

01 / Choose your connection

A way in for the way you work.

Choose by what your AI can run. The same Receiz proof objects remain the source of truth across compatible experiences.

Local AI tools

Connect the downloaded runtime to Codex, Claude Desktop, Cursor, or another host that supports local MCP. Create and verify files on your device.

Set up local tools

ChatGPT & hosted AI

Add the public HTTPS endpoint for discovery, supplied-field inspection and integration planning. Device-held execution has a separate authenticated setup.

Connect a hosted AI

AI instructions

Give an assistant the Receiz skill as project instructions or a supported skill import. Add the runtime when you want it to execute tools.

Add the instructions
02 / Local installation

Download. Connect. Create.

  1. Extract the complete package

    Download the ZIP above and extract the receiz folder into a permanent location. Keep its runtime, scripts and skills together. Install Node.js if needed, then check your terminal:

    node --version

    The major version must be 22 or higher. You do not need to run npm install.

  2. Add Receiz to your AI's MCP settings

    For hosts that accept mcpServers JSON, merge this entry into your existing configuration. Replace the argument with the absolute path to the extracted script.

    {
  "mcpServers": {
    "receiz": {
      "command": "node",
      "args": [
        "/ABSOLUTE/PATH/TO/receiz/scripts/start-mcp.mjs"
      ],
      "env": {
        "RECEIZ_MCP_DEVICE_BOOTSTRAP": "1",
        "RECEIZ_MCP_EXECUTION_PROFILE": "marketplace-nonfinancial"
      }
    }
  }
}
    Using Codex?

    Merge this into ~/.codex/config.toml, preserving your other servers. Replace the path, reconnect, and start a new conversation. Official Codex MCP instructions.

    [mcp_servers.receiz]
command = "node"
args = ["/ABSOLUTE/PATH/TO/receiz/scripts/start-mcp.mjs"]
startup_timeout_sec = 120

[mcp_servers.receiz.env]
RECEIZ_MCP_DEVICE_BOOTSTRAP = "1"
RECEIZ_MCP_EXECUTION_PROFILE = "marketplace-nonfinancial"
    Windows, paths with spaces, or a settings form?

    In JSON, use a path such as C:\\Users\\you\\Applications\\receiz\\scripts\\start-mcp.mjs. Keep the whole path in one argument. For a form: name receiz, transport stdio, command node, argument the script's absolute path, and the two environment variables above. If Node is not found, use the executable path reported by where.exe node on Windows or command -v node on macOS/Linux.

  3. Reconnect and check readiness

    Save your settings, restart the MCP connection and approve startup in your host. First launch enrolls the local device; allow up to 120 seconds where your host has a timeout setting. In a new conversation, ask:

    Use receiz_capabilities with action describe. Report the tools available on this connection. Then use receiz_offline_seal_status to report local sealing readiness. Do not create, replace or publish any proof object.

    Look for actual tool calls and returned status. The host's discovered tools define this connection; the full SDK inventory is broader.

  4. Make your first proof object

    Place a small text file in ~/.receiz/workspace, select it in your assistant, and ask:

    Seal my selected text file to a new unique output ending in .receized. Show the proposed input and output before writing. Preserve the original and do not overwrite existing files. Then independently verify the exact saved output.

    This connection exposes reviewed nonfinancial operations. Each tool retains its proof, custody and consent requirements. Settlement, Reserve transaction planning, financial transfers and unrestricted executors are excluded from this profile.

03 / Hosted connection

Bring Receiz into ChatGPT.

Use the public connection to explore Receiz and plan integrations. A hosted connection reaches the server; uploading the ZIP alone does not run the local tools.

  1. Enable developer mode

    In ChatGPT Settings, open Security and login and enable Developer mode. If the option is unavailable, check with your workspace administrator or use a local MCP host.

  2. Add the public endpoint

    Open ChatGPT Plugins, choose the plus button, and name the connection Receiz — public tools. Choose a public HTTPS connection, enter this URL, and use no authentication:

    https://receiz.com/api/mcp

  3. Review the tools and start a conversation

    Add the connection from the conversation's tools menu and ask:

    Use receiz_capabilities with action describe, then explain what this connection can actually execute.

    The public endpoint supports discovery, supplied-field inspection and planning. It does not hold your local files or signing identity, perform local sealing, or provide a canonical enclosing-artifact verification verdict from inspected fields.

Need ChatGPT to execute on your device?

The separate https://receiz.com/api/mcp/device connection composes the existing device-held runtime. It needs a running local worker, registered OAuth client, explicit account/device binding and separate caller and worker grants. The download contains no credentials. Follow the advanced section of the complete setup guide and the Receiz Connect integration. Private signing custody stays on the device.

04 / Give your assistant the instructions

Teach it how to use Receiz.

Import the marketplace-skills/receiz folder where your AI supports skills, or add its SKILL.md to your assistant's project instructions. Instructions guide your AI; the connected runtime supplies its tools.

Download assistant instructions
The next connection / Receiz Mind

Give what you learn
somewhere to grow.

Mind brings your recorded work, retained sources and sealed memory into one place to inspect and reason from. Use it alongside your AI, with the identity and source files you choose to carry.

Meet Receiz Mind Open Mind
ReceizThe proof objects you own.MindThe memory you build on.Your AIThe tools you choose to use.

Sharing is explicit. Installing the package does not automatically import Mind memory or private AI conversations.

Keep a copy

Your installation kit.

Save the guide for later, download individual configuration files, or inspect the checksum of the complete package.

Complete Receiz package

Executable tools, required dependencies and setup files.

Complete manual setup guide

Local setup, ChatGPT, device connection, identity and troubleshooting.

MCP configuration · JSON

For hosts accepting mcpServers. Replace the absolute path.

Codex configuration · TOML

Merge into your existing config.toml. Replace the absolute path.

Assistant instructions · Markdown

The scoped Receiz skill for supported skills or project instructions.

Verify the download checksum · SHA-256

Compare the downloaded ZIP with this digest using shasum -a 256 on macOS, sha256sum on Linux, or Get-FileHash -Algorithm SHA256 in PowerShell. This checks package bytes; proof object verification uses the canonical Receiz verifier.

377eeee6d32d3bc7775a6c7312277cb988222407aeb2590092d15f8f687c7173
A few useful answers

Set up once. Keep your continuity.

How do identity, files and updates work?

New local setup stores encrypted Identity Seal custody in ~/.receiz/mcp-device and files in ~/.receiz/workspace. These live outside the replaceable package. To use existing custody, configure RECEIZ_SUBJECT_IDENTITY_PATH and RECEIZ_SUBJECT_IDENTITY_PASSPHRASE privately on the host before first launch. A new default identity does not automatically become your existing account. Keep passphrases and private identity artifacts out of chat attachments.

For updates, stop the connection, extract the new package into a new folder, update the script path and reconnect. Preserve your existing custody and workspace. Keep the previous package until the new connection passes readiness.

Will this work with every AI?

Local execution requires a host that supports stdio MCP and can launch Node. Hosted execution requires a supported remote MCP connection. An AI that accepts only text can read the instructions but cannot execute Receiz tools from those instructions alone. Your AI provider's subscription and permission requirements still apply.

My assistant cannot find Node, files or tools.

Check Node 22+, the absolute script path, and that the whole ZIP was extracted. Reconnect the MCP server and start a fresh conversation. Use an absolute Node executable path if the desktop host cannot find it. Put the selected file in the configured workspace. For enrollment or custody errors, preserve the existing identity files and resolve the reported cause; do not delete custody to force a reset.

What works offline?

Local proof creation and verification work after canonical enrollment and installation of proof resources. A cloud AI still needs its own connectivity. Receiz Mind's Mac application includes local inference; remote operations and device relay require a connection.