How type.com works
Architecture, models, sandbox, credentials, access, limits, pricing, and status on one page. Each section links to the documentation behind it.
Updated
What is type.com?
type.com runs your team's automations and agents. You describe the work, and type.com writes the code and instructions, runs them with Claude or Codex on your team's shared connections, and records every run.
Teams work in Spaces, which hold their connections, skills, memory, threads, automations, documents, and apps. Each automation can combine code for the exact steps, the team's skills, a managed Claude or Codex agent for judgment, and an isolated sandbox.
Architecture
How does a type.com automation run?
A trigger starts each run: a schedule, a Slack message, a GitHub or Linear event, a webhook, an inbound email, or an RSS item. An instruction automation starts the managed agent with your instructions, the Space's skills, and its connections, and each run makes a thread with the agent's messages and tool calls.
Code automations, in early access, are TypeScript methods that type.com runs in a new QuickJS VM for each run. The VM has no fetch, file system, shell, or Node and Bun APIs, and each run gets a short-lived token rather than a provider credential. Code handles the fixed steps and hands judgment to the managed agent.
type.com doesn't make external writes exactly once, run schedule ticks missed during a pause, or keep events that arrive during a pause. Code automations use effect keys and checkpoints so a repeated run doesn't repeat its work.
Models
Which models run the agent?
Claude and Codex are the managed agents for Spaces. Each person can connect their own Claude or ChatGPT subscription, and supported requests then run through it instead of the workspace's AI usage allowance.
Requests that don't use a connected subscription are metered at the model provider's published API rates.
Sandbox
Where does agent work run?
Each automation has its own isolated Linux sandbox for shell commands, Git, the GitHub CLI, and scripts. type.com keeps it between runs for speed but can replace it at any time, so lasting state belongs in checkpoints or connected systems.
Commands in the sandbox get no type.com credential. When the Space has GitHub connections, type.com provides fresh GitHub credentials for them.
Credentials
Where do credentials live?
Connections are added once, through OAuth, API keys, service accounts, custom APIs, or hosted MCP servers, as personal or organization connections. Adding a connection to the workspace doesn't give every Space access: each connection is assigned to the Spaces that need it.
type.com stores OAuth grants and encrypted API keys on its servers. Code automation runs get a short-lived token instead of a provider credential, custom API credentials go only to the approved API host, and MCP clients receive tool metadata and results, not tokens.
Access
Who can see and change what?
Spaces are the access boundary. Everyone in a Space works with its connections, skills, memory, conversations, and Library, and a private Space is limited to the people who have been added. Channels organize work inside a Space; they aren't a separate permission boundary.
AI acts with the access of the person who started the work. Scheduled runs use the automation creator's identity, model credentials, and billing context, and admins can reassign an automation's owner.
Admins see workspace analytics, including each automation's owner, status, and AI cost. They can't open people's Sidekick conversations through Analytics.
Developer access
How do developers and AI agents work with type.com?
The type.com CLI runs on macOS, Linux, and Windows and covers apps, documents, skills, sessions, and code automations. Coding agents can follow type.com/install.md to install it and sign in.
type.com MCP gives Claude, ChatGPT, and other MCP clients OAuth access to the workspace at https://api.type.com/mcp, with the same permissions as the web app. type.com turns on MCP access for each workspace.
Every documentation page has a Markdown version, indexed at docs.type.com/llms.txt.
Limits
What are the limits?
These are the main limits for automations as of October 16, 2026. Limits can change, so check the documentation for the current list.
| Limit | Value |
|---|---|
| Run time | 180 seconds for each code automation run |
| VM memory | 1 GiB |
| Runs at the same time | 1 for each automation; other runs wait in a queue |
| Requests from code | 128 in a run, with up to 8 in progress at the same time |
| Run result | 2 MiB for output, checkpoint, report, handoffs, and action logs together |
| Sandbox | 1 for each automation; 170 seconds for one sandbox call |
| Webhook adapters | 120 incoming requests each minute for each workspace |
| Schedules | Instruction automations every 1 to 1,440 minutes, at set times, or monthly; code automations as often as once a minute |
| Failed runs | When a code automation's code fails, it pauses and starts a repair; a run that fails before its code starts is retried. A scheduled instruction automation pauses after three failed runs in a row. |
Pricing
How is type.com priced?
Plans are priced by team size: Basic is $50 a month for 2 team members and 100 credits, Pro is $100 a month for 4 team members and 200 credits, and Enterprise starts at $250 a month for 20 team members and 500 credits.
One credit is $1 of AI usage at the model provider's published rates, and requests that run on a connected Claude or ChatGPT subscription don't use credits. Every plan includes every feature and starts with a 14-day free trial.
Admins get a notice at about 80% of the monthly allowance. At 100%, new AI work pauses unless auto-upgrade is on.
PricingStatus and security
Where are status, security, and legal details?
status.type.com shows the current availability of the website and app. type.com's security documentation, controls, and subprocessors are on its trust center.
FAQ
Questions evaluators ask
- Is type.com a managed agent platform?
- Yes. type.com runs a managed Claude or Codex agent for your team's automations, with triggers, connections, a sandbox, and a thread for every run included, so there's no agent infrastructure to host.
- Does type.com store my API keys?
- type.com stores OAuth grants and encrypted API keys on its servers. Code automation runs get a short-lived token instead of a provider credential, custom API credentials go only to the approved API host, and MCP clients receive tool results, not tokens.
- Who can see an automation's runs?
- Each run makes a thread in its Space, which the Space's members can open. Admins see every automation's owner, status, and AI cost in Settings → Analytics, but they can't open people's Sidekick conversations there.
- What are the limits on a code automation run?
- A run has 180 seconds, 1 GiB of VM memory, 128 requests from code with up to 8 at a time, and 2 MiB for its combined result. Limits can change, and docs.type.com lists the current ones.
- How much does type.com cost?
- Plans start at $50 a month for 2 team members and 100 credits, where one credit is $1 of AI usage at provider rates. Connected Claude or ChatGPT subscriptions don't use credits, and every plan starts with a 14-day free trial.
- Where is type.com's security documentation?
- On the trust center at trust.type.com, which lists type.com's controls and subprocessors and answers common security questions.
Documentation behind this page
See it run on your own work.
Try type.com with your team's real connections and automations. Every plan includes every feature.
Get started for free