Token360 MCP and Agent integrations

Connect Agents to models and your enterprise control plane

Developers connect Token360 model routing with one inference key. Enterprise admins use Control MCP so trusted Agents can create sub-accounts, configure keys, and manage access within tenant boundaries.

OpenAI-compatible model APIStreamable HTTP MCPTenant-scoped authorization

Agent connection ready

Two paths

Your Agent

Agent platform

Token360

Model routing and enterprise control

01 / Unified model API

02 / Control MCP

One entryTwo key typesClear access

Choose your path

The right integration for every use case

Model inference and enterprise administration use separate credentials and permission boundaries. Pick your path and start in minutes.

Individuals and developers

Make Token360 your Agent's model layer

Create an inference API key after sign-up, then enter the Token360 endpoint in any Agent platform that supports a custom OpenAI-compatible provider.

One key for text and multimodal models across providers
Works with common OpenAI SDK and Agent configurations
Manage models, usage, and spend in one place

Create an individual account

Enterprise admins

Manage your tenant with Control MCP

Enterprise owners create a Management or Provisioning Control Key, then let trusted Agents execute organization operations through standard MCP tools.

Create, inspect, suspend, and delete sub-accounts
Create and maintain inference keys for sub-accounts
Configure model access within the current tenant

Apply for enterprise

Connection guide

Create a key, then send this prompt to your Agent

Choose your use case to get the right account entry point, a compatible Agent, and a ready-to-copy installation prompt.

01
Sign up and create an inference API key
Create an individual account, generate an inference key on the API Keys page, and store it locally as TOKEN360_API_KEY.
Create an individual account Already registered? Create an API key
02
Open a local coding Agent
Send the installation prompt to an Agent that can update local or project configuration, such as Codex, Claude Code, or Cursor Agent, not a regular web chat.
CodexClaude CodeCursor Agent
04
Run one minimal model test
Have the Agent check the endpoint and environment variable reference, then prepare a minimal request with your selected model. Review it before incurring inference cost.

Copy and send this installation prompt

The Agent will identify its host, use the appropriate official setup method, and connect without exposing your key.

Configure Token360 as the OpenAI-compatible model provider for this project.

Base URL: https://api.token360.ai/v1
Authentication: read the key from the TOKEN360_API_KEY environment variable. Do not ask me to paste the real key into chat, write it into source code, or commit it to the repository.

First identify whether the current host is Codex, Claude Code, or Cursor Agent, then use that host's official configuration method. When finished, validate the endpoint and environment variable reference, explain how to select a Token360 model, and show me the smallest test request before sending any billable inference request. Wait for my confirmation.

Do not substitute a real key into this prompt. Store it securely in the environment variable named by the prompt first.

Control MCP documentation

Connect with the Agent client you already use

Token360 hosts the remote MCP server, so there is nothing to install locally. Choose a client, copy its configuration, then list the tools to verify the connection.

Server URL: https://api.token360.ai/enterprise/mcp
Transport: Streamable HTTP
Authentication: Management or Provisioning Control Key sent as a Bearer header

Control MCP manages an enterprise tenant; it does not replace the model inference API. Applications still call https://api.token360.ai/v1 with an inference key to run models.

Copy the client configuration

The Codex CLI and IDE extension share MCP configuration; this command stores only the environment variable name.

codex mcp add token360-control \
  --url https://api.token360.ai/enterprise/mcp \
  --bearer-token-env-var TOKEN360_CONTROL_KEY

codex mcp list

Set TOKEN360_CONTROL_KEY securely on your machine first. Never put the real Control Key in this example, a chat message, or source control.

Enterprise Control MCP

Let Agents handle repetitive tenant operations

Control MCP exposes permission-aware standard tools so enterprise admins can complete common organization tasks in natural language.

Sub-account lifecycle

List, create, suspend, and reactivate enterprise members. Deletion is irreversible.

list_sub_accountscreate_sub_accountset_sub_account_statusdelete_sub_account

API key management

List, create, and maintain member inference keys. A secret is returned only once when created.

list_sub_account_api_keyscreate_sub_account_api_keyupdate_sub_account_api_keydelete_sub_account_api_key

Model access policies

Read member policy or configure inherited enterprise access and model allowlists.

get_sub_account_model_accessset_sub_account_model_access

Permissions and approval

Management Keys can cover full administration; Provisioning Keys should use least privilege. Review arguments and ask for approval before writes or deletion.

Verify with read-only toolsConfirm write toolsDeletion is irreversible

Troubleshooting

Check these three things first

Separating authentication, client configuration, and browser restrictions resolves most connection failures quickly.

The server returns 401 Unauthorized: Use an enterprise Management or Provisioning Control Key, not an inference key. Then check the Authorization Bearer header and confirm the environment variable is available to the client.
The server connects but no tools appear: Confirm the URL ends with /enterprise/mcp, restart or refresh the MCP client, then run its list or list-tools command to discover the tools again.
A browser request returns 403: Control MCP rejects direct requests carrying a browser Origin. Connect from a Streamable HTTP MCP client such as Codex, Claude Code, or Cursor.

Start now

Connect one Agent, then scale across your team

Individuals can sign up and create an inference key now. Teams that need Control MCP, sub-accounts, and governance can apply for enterprise.