Type.com launches managed platform to run shared AI agents

Moving from local to cloud

Type.com is a managed-agent platform designed to turn individual AI automations into company-wide workflows. While many people currently build agents in private accounts or run them on local laptops, Type.com provides a way to host these tasks on managed cloud machines. This shift allows a single power user to build a tool that the rest of the company can then use through Slack, Teams, or the web.

The platform specifically supports popular agent harnesses like Claude Code, Codex, and OpenCode. Instead of just sending a prompt to a model API, the system runs the full harness, which includes a terminal and a filesystem. This means the agents can perform complex tasks that require interacting with files or executing code sequences rather than just generating text.

By using a managed runtime, the platform aims to make these automations more reliable for production use. The system is designed to be self-improving, meaning it learns from every run to become more accurate and cheaper over time. This approach combines deterministic code steps with the flexible reasoning of AI agents to handle recurring business tasks.

Governance and shared access

One of the primary hurdles for companies adopting AI is managing permissions and security. Type.com addresses this by providing a central hub for credentials and tool access. It allows teams to connect to over 1,000 business tools, such as Google Drive, Meta Ads, and NetSuite, once and then share those connections across the workspace. This prevents the need for every individual employee to have their own separate login for every third-party service.

The platform offers granular permissions based on user roles, specific spaces, and tool calls. For example, a team can grant an agent the ability to read customer orders from Shopify while explicitly restricting it from performing refunds. Credentials remain encrypted and never touch the actual code, which helps maintain security standards across the organization.

This centralized model also helps with observability and auditing. Admins can see every run, catch failures, and monitor costs in real time. The system tracks inputs, steps, and outputs for every execution, and it automatically notifies administrators when AI usage reaches about 80% of the monthly allowance. This level of oversight is intended to help teams manage their budgets and ensure data safety.

Infrastructure and reliability

To ensure that agents can run reliably at scale, Type.com uses isolated environments for every task. Each agent operates in its own sandboxed virtual machine (VM) that includes a filesystem, a shell, and a browser. This isolation ensures that one run never interferes with another, which is a critical requirement for running multiple different automations simultaneously without risk of data leakage.

The platform also provides agents with a real cloud browser that can use saved logins or shared logins from a team vault. This allows agents to interact with web-based tools just as a human would. Because the runs are durable, the system uses checkpointing to ensure that if a timeout or a flaky API occurs, the work is not lost. The system will automatically retry the failed steps to complete the task.

Users can trigger these automations through various methods, including set schedules, webhooks, or specific events in connected tools. This flexibility allows for a wide range of use cases, from a daily sales report that runs every morning at 8:00 AM to a reactive workflow that triggers whenever a specific change occurs in a connected app. The platform handles the underlying infrastructure so users don't have to manage servers or cron jobs.

Deployment and pricing

Type.com is designed to integrate with the tools that teams already use for development. Users can start building agents in Claude, Codex, or ChatGPT and then deploy them to the managed cloud in one step using built-in plugins. This workflow is intended to make it easy for developers to move from a prototype on their own machine to a production-ready tool that the whole team can access.

The platform offers different ways to handle the underlying AI costs. While there are paid plans starting at $50 per month for two team members, users also have the option to run requests through their own existing Claude or ChatGPT subscriptions. This allows companies to leverage their current contracts while still gaining the benefits of the managed orchestration and governance layer.

The service is positioned as a way to solve the problem of 'siloed' AI. When useful agents are stuck in one person's account or on a single laptop, the rest of the team cannot benefit from them. Type.com provides the infrastructure to move those tasks into a shared space where they can be monitored, governed, and scaled across marketing, sales, and operations departments.