Docs
Hosted agent platform

Ship cloud agents. Skip the infrastructure.

Hosted runtime in isolated sandboxes, with all the building blocks you need to run production-ready agents.

Get started for freeRead the docs
ClaudeOpenAIGeminiGrokMetaKimi+50 models

No fees, pay only official API prices.

Deploy a production-ready agent in seconds

pip install m8tes · npm i @m8tes/sdk · npm i @m8tes/react

Example run: $ python quickstart.py. Output: Hi Maya — circling back on Acme's renewal. Happy to send SSO pricing and lock a Q3 kickoff. Tue or Wed work for a quick call?. run completed · 2.1s · $0.005.

example run · run completed · 2.1s · $0.005
from m8tes import M8tes

client = M8tes()  # reads M8TES_API_KEY
# Personal development only: this disables strict user_id checks account-wide.
# This persists. For customer-facing apps, keep strict mode on and pass user_id.
client.settings.update(require_end_user_id=False)

# Connect a model provider first: https://m8tes.ai/account
for chunk in client.runs.stream_text(
    message="Draft a follow-up for Acme's stalled renewal — SSO pricing and a Q3 start.",
    model="grok-4.6",  # match whatever provider you connected
    raise_on_error=True,
):
    print(chunk, end="", flush=True)

All the building blocks to ship cloud agents in seconds

Isolated sandbox per run

190+ integrations

Memory across runs

Persistent filesystem

Email, Slack & webhooks

Human approvals

Webhook & app events

Scheduled runs

Desktop & browser

Agents manage agents

Each user gets their own agent inside your product

yourapp.com/assistantEmbedded with @m8tes/react
Assistantuser_id: cus_4821

Example: the user asks "which accounts need attention before Friday?" and the agent replies "Acme and Northwind. Acme's renewal is blocked on SSO; Northwind's proposal expires Friday. Want me to draft both follow-ups?". streamed into your product · user_id cus_4821.

example run · streamed into your product · user_id cus_4821

Need help from an engineer?

Book a call

FAQ

Customer API runs are billed per token at the model provider’s list price, with no token markup. Usage comes from your prepaid balance. A run that fails after using tokens still incurs that usage.

Billing and usage docs

New accounts start with a zero prepaid balance. Add funds before running an example with a customer user_id. Your Platform plan or personal model subscription does not fund these customer runs.

Yes. Call the API from your server using Python, TypeScript, or HTTP. Use a stable user_id for each customer’s runs, memory, and app connections. Your application handles sign-in, the interface, and mapping each customer to the right user_id.

Customer isolation docs

Start a connection for the customer’s user_id. The app determines the authorization method: OAuth, an API key, or another supported flow. The agent’s available actions depend on the integration and the permissions the customer grants.

Integration docs

Auto mode pauses for actions that need permission. Plan mode asks for a plan review first. Routine permitted work and actions covered by saved rules can continue without another prompt. Autonomous mode skips tool-approval prompts. Your application can show approval requests and return the user’s answer.

Approval docs

Metadata-only mode limits stored conversation, tool, and report content for new runs. Operational metadata and configuration, including starting instructions, remain. Earlier runs are not deleted. This limits history, report retrieval, and new memory. Provider retention is separate; check the model and funding path.

Data retention docs